Chapter 15. MySQL Cluster

Table of Contents

15.1. MySQL Cluster Overview
15.2. Basic MySQL Cluster Concepts
15.2.1. MySQL Cluster Nodes, Node Groups, Replicas, and Partitions
15.3. Simple Multi-Computer How-To
15.3.1. Hardware, Software, and Networking
15.3.2. Multi-Computer Installation
15.3.3. Multi-Computer Configuration
15.3.4. Initial Startup
15.3.5. Loading Sample Data and Performing Queries
15.3.6. Safe Shutdown and Restart
15.4. MySQL Cluster Configuration
15.4.1. Building MySQL Cluster from Source Code
15.4.2. Installing the Software
15.4.3. Quick Test Setup of MySQL Cluster
15.4.4. Configuration File
15.4.5. Overview of Cluster Configuration Parameters
15.4.6. Configuring Parameters for Local Checkpoints
15.5. Upgrading and Downgrading MySQL Cluster
15.5.1. Performing a Rolling Restart of the Cluster
15.5.2. Cluster Upgrade and Downgrade Compatibility
15.6. Process Management in MySQL Cluster
15.6.1. MySQL Server Process Usage for MySQL Cluster
15.6.2. ndbd, the Storage Engine Node Process
15.6.3. ndb_mgmd, the Management Server Process
15.6.4. ndb_mgm, the Management Client Process
15.6.5. Command Options for MySQL Cluster Processes
15.7. Management of MySQL Cluster
15.7.1. MySQL Cluster Startup Phases
15.7.2. Commands in the Management Client
15.7.3. Event Reports Generated in MySQL Cluster
15.7.4. Single User Mode
15.8. On-line Backup of MySQL Cluster
15.8.1. Cluster Backup Concepts
15.8.2. Using The Management Client to Create a Backup
15.8.3. How to Restore a Cluster Backup
15.8.4. Configuration for Cluster Backup
15.8.5. Backup Troubleshooting
15.9. Cluster Utility Programs
15.9.1. ndb_config
15.9.2. ndb_delete_all
15.9.3. ndb_desc
15.9.4. ndb_drop_index
15.9.5. ndb_drop_table
15.9.6. ndb_error_reporter
15.9.7. ndb_print_backup_file
15.9.8. ndb_print_schema_file
15.9.9. ndb_print_sys_file
15.9.10. ndb_select_all
15.9.11. ndb_select_count
15.9.12. ndb_show_tables
15.9.13. — NDBCluster Size Requirement Estimator
15.9.14. ndb_waiter
15.10. MySQL Cluster Replication
15.10.1. Abbreviations and Symbols
15.10.2. Assumptions and General Requirements
15.10.3. Known Issues
15.10.4. Replication Schema and Tables
15.10.5. Preparing the Cluster for Replication
15.10.6. Starting Replication (Single Replication Channel)
15.10.7. Using Two Replication Channels
15.10.8. Implementing Failover with MySQL Cluster
15.10.9. MySQL Cluster Backups With Replication
15.11. MySQL Cluster Disk Data Storage
15.12. Using High-Speed Interconnects with MySQL Cluster
15.12.1. Configuring MySQL Cluster to use SCI Sockets
15.12.2. Understanding the Impact of Cluster Interconnects
15.13. Known Limitations of MySQL Cluster
15.14. MySQL Cluster Development Roadmap
15.14.1. MySQL Cluster Changes in MySQL 5.1
15.15. MySQL Cluster Glossary

Due to a build slippage, binary distributions of MySQL 5.1.12 contained neither NDB Cluster nor Partitioning. We apologize for any inconvenience. Please upgrade to 5.1.14. if you build from source, you can execute configure with the --with-ndbcluster and --with-partition options.

MySQL Cluster is a high-availability, high-redundancy version of MySQL adapted for the distributed computing environment. It uses the NDB Cluster storage engine to enable running several MySQL servers in a cluster. This storage engine is available in MySQL 5.1 binary releases and in RPMs compatible with most modern Linux distributions.

MySQL Cluster is currently available and supported on the following platforms:

For exact levels of support available for MySQL Cluster on specific combinations of operating system versions, operating system distributions, and hardware platforms, please refer to the Cluster Supported Platforms list maintained by the MySQL Support Team on the MySQL AB Web site.

MySQL Cluster is not currently supported on Microsoft Windows. We are working to make Cluster available on all operating systems supported by MySQL, including Windows, and will update the information provided here as this work continues.

This chapter represents a work in progress, and its contents are subject to revision as MySQL Cluster continues to evolve. Additional information regarding MySQL Cluster can be found on the MySQL AB Web site at

Additional resources

15.1. MySQL Cluster Overview

MySQL Cluster is a technology that enables clustering of in-memory databases in a shared-nothing system. The shared-nothing architecture allows the system to work with very inexpensive hardware, and with a minimum of specific requirements for hardware or software.

MySQL Cluster is designed not to have any single point of failure. For this reason, each component is expected to have its own memory and disk, and the use of shared storage mechanisms such as network shares, network filesystems, and SANs is not recommended or supported.

MySQL Cluster integrates the standard MySQL server with an in-memory clustered storage engine called NDB. In our documentation, the term NDB refers to the part of the setup that is specific to the storage engine, whereas “MySQL Cluster” refers to the combination of MySQL and the NDB storage engine.

A MySQL Cluster consists of a set of computers, each running a one or more processes which may include a MySQL server, a data node, a management server, and (possibly) a specialized data access programs. The relationship of these components in a cluster is shown here:

MySQL Cluster Components

All these programs work together to form a MySQL Cluster. When data is stored in the NDB Cluster storage engine, the tables are stored in the data nodes. Such tables are directly accessible from all other MySQL servers in the cluster. Thus, in a payroll application storing data in a cluster, if one application updates the salary of an employee, all other MySQL servers that query this data can see this change immediately.

The data stored in the data nodes for MySQL Cluster can be mirrored; the cluster can handle failures of individual data nodes with no other impact than that a small number of transactions are aborted due to losing the transaction state. Because transactional applications are expected to handle transaction failure, this should not be a source of problems.

15.2. Basic MySQL Cluster Concepts

NDB is an in-memory storage engine offering high-availability and data-persistence features.

The NDB storage engine can be configured with a range of failover and load-balancing options, but it is easiest to start with the storage engine at the cluster level. MySQL Cluster's NDB storage engine contains a complete set of data, dependent only on other data within the cluster itself.

The cluster portion of MySQL Cluster is currently configured independently of the MySQL servers. In a MySQL Cluster, each part of the cluster is considered to be a node.

Note: In many contexts, the term “node” is used to indicate a computer, but when discussing MySQL Cluster it means a process. It is possible to run any number of nodes on a single computer, for which we use the term cluster host.

(However, it should be noted MySQL does not currently support the use of multiple data nodes on a single computer in a production setting. See Issues exclusive to MySQL Cluster.)

There are three types of cluster nodes, and in a minimal MySQL Cluster configuration, there will be at least three nodes, one of each of these types:

  • Management node (MGM node): The role of this type of node is to manage the other nodes within the MySQL Cluster, performing such functions as providing configuration data, starting and stopping nodes, running backup, and so forth. Because this node type manages the configuration of the other nodes, a node of this type should be started first, before any other node. An MGM node is started with the command ndb_mgmd.

  • Data node: This type of node stores cluster data. There are as many data nodes as there are replicas, times the number of fragments. For example, with two replicas, each having two fragments, you will need four data nodes. It is not necessary to have more than one replica. A data node is started with the command ndbd.

  • SQL node: This is a node that accesses the cluster data. In the case of MySQL Cluster, an SQL node is a traditional MySQL server that uses the NDB Cluster storage engine. An SQL node is typically started with the command mysqld --ndbcluster or by using mysqld with the ndbcluster option added to my.cnf.

    An SQL node is actually just a specialised type of API node, which designates any application which accesses Cluster data. One example of an API node is the ndb_restore utility that is used to restore a cluster backup. It is possible to write such applications using the NDB API.

Important: It is not realistic to expect to employ a three-node setup in a production environment. Such a configuration provides no redundancy; in order to benefit from MySQL Cluster's high-availability features, you must use multiple data and SQL nodes. The use of multiple management nodes is also highly recommended.

For a brief introduction to the relationships between nodes, node groups, replicas, and partitions in MySQL Cluster, see Section 15.2.1, “MySQL Cluster Nodes, Node Groups, Replicas, and Partitions”.

Configuration of a cluster involves configuring each individual node in the cluster and setting up individual communication links between nodes. MySQL Cluster is currently designed with the intention that data nodes are homogeneous in terms of processor power, memory space, and bandwidth. In addition, to provide a single point of configuration, all configuration data for the cluster as a whole is located in one configuration file.

The management server (MGM node) manages the cluster configuration file and the cluster log. Each node in the cluster retrieves the configuration data from the management server, and so requires a way to determine where the management server resides. When interesting events occur in the data nodes, the nodes transfer information about these events to the management server, which then writes the information to the cluster log.

In addition, there can be any number of cluster client processes or applications. These are of two types:

  • Standard MySQL clients: These are no different for MySQL Cluster than they are for standard (non-Cluster) MySQL. In other words, MySQL Cluster can be accessed from existing MySQL applications written in PHP, Perl, C, C++, Java, Python, Ruby, and so on.

  • Management clients: These clients connect to the management server and provide commands for starting and stopping nodes gracefully, starting and stopping message tracing (debug versions only), showing node versions and status, starting and stopping backups, and so on.

15.2.1. MySQL Cluster Nodes, Node Groups, Replicas, and Partitions

This section discusses the manner in which MySQL Cluster divides and duplicates data for storage.

Central to an understanding of this topic are the following concepts, listed here with brief definitions:

  • (Data) Node: An ndbd process, which stores a replica —that is, a copy of the partition (see below) assigned to the node group of which the node is a member.

    Each data node should be located on a separate computer. While it is also possible to host multiple ndbd processes on a single computer, such a configuration is not supported.

    It is common for the terms “node” and “data node” to be used interchangeably when referring to an ndbd process; where mentioned, management (MGM) nodes (ndb_mgmd processes) and SQL nodes (mysqld processes) are specified as such in this discussion.

  • Node Group: A node group consists of one or more nodes, and stores partitions, or sets of replicas (see next item).

    Note: Currently, all node groups in a cluster must have the same number of nodes.

  • Partition: This is a portion of the data stored by the cluster. There are as many cluster partitions as nodes participating in the cluster. Each node is responsible for keeping at least one copy of any partitions assigned to it (that is, at least one replica) available to the cluster.

    A replica belongs entirely to a single node; a node can (and usually does) store several replicas.

  • Replica: This is a copy of a cluster partition. Each node in a node group stores a replica. Also sometimes known as a partition replica. The number of replicas is equal to the number of nodes per node group.

The following diagram illustrates a MySQL Cluster with four data nodes, arranged in two node groups of two nodes each; nodes 1 and 2 belong to node group 0, and nodes 3 and 4 belong to node group 1. Note that only data (ndbd) nodes are shown here; although a working cluster requires an ndb_mgm process for cluster management and at least one SQL node to access the data stored by the cluster, these have been omitted in the figure for clarity.

A MySQL Cluster, with 2 node groups having 2
          nodes each

The data stored by the cluster is divided into four partitions, numbered 0, 1, 2, and 3. Each partition is stored — in multiple copies — on the same node group. Partitions are stored on alternate node groups:

  • Partition 0 is stored on node group 0; a primary replica (primary copy) is stored on node 1, and a backup replica (backup copy of the partition) is stored on node 2.

  • Partition 1 is stored on the other node group (node group 1); this partition's primary replica is on node 3, and its backup replica is on node 4.

  • Partition 2 is stored on node group 0. However, the placing of its two replicas is reversed from that of Partition 0; for Partition 2, the primary replica is stored on node 2, and the backup on node 1.

  • Partition 3 is stored on node group 1, and the placement of its two replicas are reversed from those of partition 1. That is, its primary replica is located on node 4, with the backup on node 3.

What this means regarding the continued operation of a MySQL Cluster is this: so long as each node group participating in the cluster has at least one node operating, the cluster has a complete copy of all data and remains viable. This is illustrated in the next diagram.

Nodes required to keep a 2x2 cluster

In this example, where the cluster consists of two node groups of two nodes each, any combination of at least one node in node group 0 and at least one node in node group 1 is sufficient to keep the cluster “alive” (indicated by arrows in the diagram). However, if both nodes from either node group fail, the remaining two nodes are not sufficient (shown by the arrows marked out with an X); in either case, the cluster has lost an entire partition and so can no longer provide access to a complete set of all cluster data.

15.3. Simple Multi-Computer How-To

This section is a “How-To” that describes the basics for how to plan, install, configure, and run a MySQL Cluster. Whereas the examples in Section 15.4, “MySQL Cluster Configuration” provide more in-depth information on a variety of clustering options and configuration, the result of following the guidelines and procedures outlined here should be a usable MySQL Cluster which meets the minimum requirements for availability and safeguarding of data.

This section covers hardware and software requirements; networking issues; installation of MySQL Cluster; configuration issues; starting, stopping, and restarting the cluster; loading of a sample database; and performing queries.

Basic Assumptions

This How-To makes the following assumptions:

  1. The cluster setup has four nodes, each on a separate host, and each with a fixed network address on a typical Ethernet as shown here:

    NodeIP Address
    Management (MGM) node192.168.0.10
    MySQL server (SQL) node192.168.0.20
    Data (NDBD) node "A"
    Data (NDBD) node "B"

    This may be made clearer in the following diagram:

    MySQL Cluster Multi-Computer

    Note: In the interest of simplicity (and reliability), this How-To uses only numeric IP addresses. However, if DNS resolution is available on your network, it is possible to use hostnames in lieu of IP addresses in configuring Cluster. Alternatively, you can use the /etc/hosts file or your operating system's equivalent for providing a means to do host lookup if such is available.

  2. Each host in our scenario is an Intel-based desktop PC running a common, generic Linux distribution installed to disk in a standard configuration, and running no unnecessary services. The core OS with standard TCP/IP networking capabilities should be sufficient. Also for the sake of simplicity, we also assume that the filesystems on all hosts are set up identically. In the event that they are not, you will need to adapt these instructions accordingly.

  3. Standard 100 Mbps or 1 gigabit Ethernet cards are installed on each machine, along with the proper drivers for the cards, and that all four hosts are connected via a standard-issue Ethernet networking appliance such as a switch. (All machines should use network cards with the same throughout. That is, all four machines in the cluster should have 100 Mbps cards or all four machines should have 1 Gbps cards.) MySQL Cluster will work in a 100 Mbps network; however, gigabit Ethernet will provide better performance.

    Note that MySQL Cluster is not intended for use in a network for which throughput is less than 100 Mbps. For this reason (among others), attempting to run a MySQL Cluster over a public network such as the Internet is not likely to be successful, and is not recommended.

  4. For our sample data, we will use the world database which is available for download from the MySQL AB Web site. As this database takes up a relatively small amount of space, we assume that each machine has 256MB RAM, which should be sufficient for running the operating system, host NDB process, and (for the data nodes) for storing the database.

Although we refer to a Linux operating system in this How-To, the instructions and procedures that we provide here should be easily adaptable to other supported operating systems. We also assume that you already know how to perform a minimal installation and configuration of the operating system with networking capability, or that you are able to obtain assistance in this elsewhere if needed.

We discuss MySQL Cluster hardware, software, and networking requirements in somewhat greater detail in the next section. (See Section 15.3.1, “Hardware, Software, and Networking”.)

15.3.1. Hardware, Software, and Networking

One of the strengths of MySQL Cluster is that it can be run on commodity hardware and has no unusual requirements in this regard, other than for large amounts of RAM, due to the fact that all live data storage is done in memory. (Note that this is not the case with Disk Data tables — see Section 15.11, “MySQL Cluster Disk Data Storage”, for more information about these.) Naturally, multiple and faster CPUs will enhance performance. Memory requirements for other Cluster processes are relatively small.

The software requirements for Cluster are also modest. Host operating systems do not require any unusual modules, services, applications, or configuration to support MySQL Cluster. For supported operating systems, a standard installation should be sufficient. The MySQL software requirements are simple: all that is needed is a production release of MySQL 5.1 to have Cluster support. It is not necessary to compile MySQL yourself merely to be able to use Cluster. In this How-To, we assume that you are using the server binary appropriate to your operating system, available via the MySQL software downloads page at

For inter-node communication, Cluster supports TCP/IP networking in any standard topology, and the minimum expected for each host is a standard 100 Mbps Ethernet card, plus a switch, hub, or router to provide network connectivity for the cluster as a whole. We strongly recommend that a MySQL Cluster be run on its own subnet which is not shared with non-Cluster machines for the following reasons:

  • Security: Communications between Cluster nodes are not encrypted or shielded in any way. The only means of protecting transmissions within a MySQL Cluster is to run your Cluster on a protected network. If you intend to use MySQL Cluster for Web applications, the cluster should definitely reside behind your firewall and not in your network's De-Militarized Zone (DMZ) or elsewhere.

  • Efficiency: Setting up a MySQL Cluster on a private or protected network allows the cluster to make exclusive use of bandwidth between cluster hosts. Using a separate switch for your MySQL Cluster not only helps protect against unauthorized access to Cluster data, it also ensures that Cluster nodes are shielded from interference caused by transmissions between other computers on the network. For enhanced reliability, you can use dual switches and dual cards to remove the network as a single point of failure; many device drivers support failover for such communication links.

It is also possible to use the high-speed Scalable Coherent Interface (SCI) with MySQL Cluster, but this is not a requirement. See Section 15.12, “Using High-Speed Interconnects with MySQL Cluster”, for more about this protocol and its use with MySQL Cluster.

15.3.2. Multi-Computer Installation

Each MySQL Cluster host computer running data or SQL nodes must have installed on it a MySQL server binary. For management nodes, it is not necessary to install the MySQL server binary, but you do have to install the MGM server daemon and client binaries (ndb_mgmd and ndb_mgm, respectively). This section covers the steps necessary to install the correct binaries for each type of Cluster node.

MySQL AB provides precompiled binaries that support Cluster, and there is generally no need to compile these yourself. Therefore, the first step in the installation process for each cluster host is to download the file mysql-5.1.15-beta-pc-linux-gnu-i686.tar.gz from the MySQL downloads area. We assume that you have placed it in each machine's /var/tmp directory. (If you do require a custom binary, see Section 2.9.3, “Installing from the Development Source Tree”.)

RPMs are also available for both 32-bit and 64-bit Linux platforms. For a MySQL Cluster, four RPMs are required:

  • The Server RPM (for example, MySQL-server-5.1.15-beta-0.glibc23.i386.rpm), which supplies the core files needed to run a MySQL Server.

  • The Server/Max RPM (for example, MySQL-Max-5.1.15-beta-0.glibc23.i386.rpm), which provides a MySQL Server binary with clustering support.

  • The NDB Cluster - Storage engine RPM (for example, MySQL-ndb-storage-5.1.15-beta-0.glibc23.i386.rpm), which supplies the MySQL Cluster data node binary (ndbd).

  • The NDB Cluster - Storage engine management RPM (for example, MySQL-ndb-management-5.1.15-beta-0.glibc23.i386.rpm), which provides the MySQL Cluster management server binary (ndb_mgmd).

In addition, you should also obtain the NDB Cluster - Storage engine basic tools RPM (for example, MySQL-ndb-tools-5.1.15-beta-0.glibc23.i386.rpm), which supplies several useful applications for working with a MySQL Cluster. The most important of the these is the MySQL Cluster management client (ndb_mgm). The NDB Cluster - Storage engine extra tools RPM (for example, MySQL-ndb-extra-5.1.15-beta-0.glibc23.i386.rpm) contains some additional testing and monitoring programs, but is not required to install a MySQL Cluster. (For more information about these additional programs, see Section 15.9, “Cluster Utility Programs”.)

The MySQL version number in the RPM filenames (shown here as 5.1.15-beta) can vary according to the version which you are actually using. It is very important that all of the Cluster RPMs to be installed have the same MySQL version number. The glibc version number (if present — shown here as glibc23), and architecture designation (shown here as i386) should be appropriate to the machine on which the RPM is to be installed.

See Section 2.4, “Installing MySQL on Linux”, for general information about installing MySQL using RPMs supplied by MySQL AB.

After installing from RPM, you still need to configure the cluster as discussed in Section 15.3.3, “Multi-Computer Configuration”.

Note: After completing the installation, do not yet start any of the binaries. We show you how to do so following the configuration of all nodes.

Data and SQL Node Installation — .tar.gz Binary

On each of the machines designated to host data or SQL nodes, perform the following steps as the system root user:

  1. Check your /etc/passwd and /etc/group files (or use whatever tools are provided by your operating system for managing users and groups) to see whether there is already a mysql group and mysql user on the system. Some OS distributions create these as part of the operating system installation process. If they are not already present, create a new mysql user group, and then add a mysql user to this group:

    shell> groupadd mysql
    shell> useradd -g mysql mysql

    The syntax for useradd and groupadd may differ slightly on different versions of Unix, or they may have different names such as adduser and addgroup.

  2. Change location to the directory containing the downloaded file, unpack the archive, and create a symlink to the mysql. Note that the actual file and directory names will vary according to the MySQL version number.

    shell> cd /var/tmp
    shell> tar -C /usr/local -xzvf mysql-5.1.15-beta-pc-linux-gnu-i686.tar.gz
    shell> ln -s /usr/local/mysql-5.1.15-beta-pc-linux-gnu-i686 /usr/local/mysql
  3. Change location to the mysql directory and run the supplied script for creating the system databases:

    shell> cd mysql
    shell> scripts/mysql_install_db --user=mysql
  4. Set the necessary permissions for the MySQL server and data directories:

    shell> chown -R root .
    shell> chown -R mysql data
    shell> chgrp -R mysql .

    Note that the data directory on each machine hosting a data node is /usr/local/mysql/data. We will use this piece of information when we configure the management node. (See Section 15.3.3, “Multi-Computer Configuration”.)

  5. Copy the MySQL startup script to the appropriate directory, make it executable, and set it to start when the operating system is booted up:

    shell> cp support-files/mysql.server /etc/rc.d/init.d/
    shell> chmod +x /etc/rc.d/init.d/mysql.server
    shell> chkconfig --add mysql.server

    (The startup scripts directory may vary depending on your operating system and version — for example, in some Linux distributions, it is /etc/init.d.)

    Here we use Red Hat's chkconfig for creating links to the startup scripts; use whatever means is appropriate for this purpose on your operating system and distribution, such as update-rc.d on Debian.

Remember that the preceding steps must be performed separately on each machine where a data node or SQL node is to reside.

SQL Node Installation — RPM Files

On each machine to be used for hosting a cluster SQL node, install the MySQL Max RPM by executing the following command as the system root user, replacing the name shown for the RPM as necessary to match the name of the RPM downloaded from the MySQL AB web site:

shell> rpm -Uhv MySQL-server-5.1.15-beta-0.glibc23.i386.rpm
shell> rpm -Uhv MySQL-Max-5.1.15-beta-0.glibc23.i386.rpm

This installs the MySQL server binary (mysqld) in the /usr/sbin directory, as well as all needed MySQL Server support files. It also installs the mysql.server and mysqld_safe startup scripts in /usr/share/mysql and /usr/bin, respectively. The RPM installer should take care of general configuration issues (such as creating the mysql user and group, if needed) automatically.

Data Node Installation — RPM Files

On a computer that is to host a cluster data node it is necessary to install only the NDB Cluster - Storage engine RPM. To do so, copy this RPM to the data node host, and run the following command as the system root user, replacing the name shown for the RPM as necessary to match that of the RPM downloaded from the MySQL AB web site:

shell> rpm -Uhv MySQL-ndb-storage-5.1.15-beta-0.glibc23.i386.rpm

The previous command installs the MySQL Cluster data node binary (ndbd) in the /usr/sbin directory.

Management Node Installation — .tar.gz Binary

Installation for the management (MGM) node does not require installation of the mysqld binary. Only the binaries for the MGM server and client are required, which can be found in the downloaded archive. Again, we assume that you have placed this file in /var/tmp.

As system root (that is, after using sudo, su root, or your system's equivalent for temporarily assuming the system administrator account's privileges), perform the following steps to install ndb_mgmd and ndb_mgm on the Cluster management node host:

  1. Change location to the /var/tmp directory, and extract the ndb_mgm and ndb_mgmd from the archive into a suitable directory such as /usr/local/bin:

    shell> cd /var/tmp
    shell> tar -zxvf mysql-5.1.15-beta-pc-linux-gnu-i686.tar.gz
    shell> cd mysql-5.1.15-beta-pc-linux-gnu-i686
    shell> cp /bin/ndb_mgm* /usr/local/bin

    (You can safely delete the directory created by unpacking the downloaded archive, and the files it contains, from /var/tmp once ndb_mgm and ndb_mgmd have been copied to the executables directory.)

  2. Change location to the directory into which you copied the files, and then make both of them executable:

    shell> cd /usr/local/bin
    shell> chmod +x ndb_mgm*

Management Node Installation — RPM File

To install the MySQL Cluster management server, it is necessary only to use the NDB Cluster - Storage engine management RPM. Copy this RPM to the computer intended to host the management node, and then install it by running the following command as the system root user (replace the name shown for the RPM as necessary to match that of the Storage engine management RPM downloaded from the MySQL AB web site):

shell> rpm -Uhv MySQL-ndb-management-5.1.15-beta-0.glibc23.i386.rpm

This installs the management server binary (ndb_mgmd) to the /usr/sbin directory.

You should also install the NDB management client, which is supplied by the Storage engine basic tools RPM. Copy this RPM to the same computer as the management node, and then install it by running the following command as the system root user (again, replace the name shown for the RPM as necessary to match that of the Storage engine basic tools RPM downloaded from the MySQL AB web site):

shell> rpm -Uhv MySQL-ndb-tools-5.1.15-beta-0.glibc23.i386.rpm

The Storage engine basic tools RPM installs the MySQL Cluster management client (ndb_mgm) to the /usr/bin directory.

In Section 15.3.3, “Multi-Computer Configuration”, we create configuration files for all of the nodes in our example Cluster.

15.3.3. Multi-Computer Configuration

For our four-node, four-host MySQL Cluster, it is necessary to write four configuration files, one per node host.

  • Each data node or SQL node requires a my.cnf file that provides two pieces of information: a connectstring that tells the node where to find the MGM node, and a line telling the MySQL server on this host (the machine hosting the data node) to run in NDB mode.

    For more information on connectstrings, see Section, “The Cluster connectstring.

  • The management node needs a config.ini file telling it how many replicas to maintain, how much memory to allocate for data and indexes on each data node, where to find the data nodes, where to save data to disk on each data node, and where to find any SQL nodes.

Configuring the Storage and SQL Nodes

The my.cnf file needed for the data nodes is fairly simple. The configuration file should be located in the /etc directory and can be edited using any text editor. (Create the file if it does not exist.) For example:

shell> vi /etc/my.cnf

We show vi being used here to create the file, but any text editor should work just as well.

For each data node and SQL node in our example setup, my.cnf should look like this:

# Options for mysqld process:
ndbcluster                      # run NDB storage engine
ndb-connectstring=  # location of management server

# Options for ndbd process:
ndb-connectstring=  # location of management server

After entering the preceding information, save this file and exit the text editor. Do this for the machines hosting data node “A”, data node “B”, and the SQL node.

Important: Once you have started a mysqld process with the ndbcluster and ndb-connectstring parameters in the [MYSQLD] in the my.cnf file as shown previously, you cannot execute any CREATE TABLE or ALTER TABLE statements without having actually started the cluster. Otherwise, these statements will fail with an error. This is by design.

Configuring the Management Node

The first step in configuring the MGM node is to create the directory in which the configuration file can be found and then to create the file itself. For example (running as root):

shell> mkdir /var/lib/mysql-cluster
shell> cd /var/lib/mysql-cluster
shell> vi config.ini

For our representative setup, the config.ini file should read as follows:

# Options affecting ndbd processes on all data nodes:
NoOfReplicas=2    # Number of replicas
DataMemory=80M    # How much memory to allocate for data storage
IndexMemory=18M   # How much memory to allocate for index storage
                  # For DataMemory and IndexMemory, we have used the
                  # default values. Since the "world" database takes up
                  # only about 500KB, this should be more than enough for
                  # this example Cluster setup.

# TCP/IP options:
portnumber=2202   # This the default; however, you can use any
                  # port that is free for all the hosts in cluster
                  # Note: It is recommended beginning with MySQL 5.0 that
                  # you do not specify the portnumber at all and simply allow
                  # the default value to be used instead

# Management process options:
hostname=           # Hostname or IP address of MGM node
datadir=/var/lib/mysql-cluster  # Directory for MGM node log files

# Options for data node "A":
                                # (one [NDBD] section per data node)
hostname=           # Hostname or IP address
datadir=/usr/local/mysql/data   # Directory for this data node's data files

# Options for data node "B":
hostname=           # Hostname or IP address
datadir=/usr/local/mysql/data   # Directory for this data node's data files

# SQL node options:
hostname=           # Hostname or IP address
                                # (additional mysqld connections can be
                                # specified for this node for various
                                # purposes such as running ndb_restore)

(Note: The world database can be downloaded from, where it can be found listed under “Examples”.)

After all the configuration files have been created and these minimal options have been specified, you are ready to proceed with starting the cluster and verifying that all processes are running. We discuss how this is done in Section 15.3.4, “Initial Startup”.

For more detailed information about the available MySQL Cluster configuration parameters and their uses, see Section 15.4.4, “Configuration File”, and Section 15.4, “MySQL Cluster Configuration”. For configuration of MySQL Cluster as relates to making backups, see Section 15.8.4, “Configuration for Cluster Backup”.

Note: The default port for Cluster management nodes is 1186; the default port for data nodes is 2202. Beginning with MySQL 5.0.3, this restriction is lifted, and the cluster automatically allocates ports for data nodes from those that are already free.

15.3.4. Initial Startup

Starting the cluster is not very difficult after it has been configured. Each cluster node process must be started separately, and on the host where it resides. The management node ahould be started first, followed by the data nodes, and then finally by any SQL nodes:

  1. On the management host, issue the following command from the system shell to start the MGM node process:

    shell> ndb_mgmd -f /var/lib/mysql-cluster/config.ini

    Note that ndb_mgmd must be told where to find its configuration file, using the -f or --config-file option. (See Section 15.6.3, “ndb_mgmd, the Management Server Process”, for details.)

  2. On each of the data node hosts, run this command to start the ndbd process for the first time:

    shell> ndbd --initial

    Note that it is very important to use the --initial parameter only when starting ndbd for the first time, or when restarting after a backup/restore operation or a configuration change. This is because the --initial option causes the node to delete any files created by earlier ndbd instances that are needed for recovery, including the recovery log files.

    An exception to this is that --initial does not delete Disk Data files. If you do need to perform an initial restart of the cluster, you must delete any existing Disk Data log files and data files manually.

  3. If you used RPM files to install MySQL on the cluster host where the SQL node is to reside, you can (and should) use the supplied startup script to start the MySQL server process on the SQL node.

If all has gone well, and the cluster has been set up correctly, the cluster should now be operational. You can test this by invoking the ndb_mgm management node client. The output should look like that shown here, although you might see some slight differences in the output depending upon the exact version of MySQL that you are using:

shell> ndb_mgm
-- NDB Cluster -- Management Client --
ndb_mgm> SHOW
Connected to Management Server at: localhost:1186
Cluster Configuration
[ndbd(NDB)]     2 node(s)
id=2    @  (Version: 5.1.15-beta, Nodegroup: 0, Master)
id=3    @  (Version: 5.1.15-beta, Nodegroup: 0)

[ndb_mgmd(MGM)] 1 node(s)
id=1    @  (Version: 5.1.15-beta)

[mysqld(SQL)]   1 node(s)
id=4   (Version: 5.1.15-beta)

Note: The SQL node is referenced here as [mysqld(API)]. This is perfectly normal, and reflects the fact that the mysqld process is acting as a cluster API node.

You should now be ready to work with databases, tables, and data in MySQL Cluster. See Section 15.3.5, “Loading Sample Data and Performing Queries”, for a brief discussion.

15.3.5. Loading Sample Data and Performing Queries

Working with data in MySQL Cluster is not much different from doing so in MySQL without Cluster. There are two points to keep in mind:

  • For a table to be replicated in the cluster, it must use the NDB Cluster storage engine. To specify this, use the ENGINE=NDB or ENGINE=NDBCLUSTER table option. You can add this option when creating the table:


    Alternatively, for an existing table that uses a different storage engine, use ALTER TABLE to change the table to use NDB Cluster:

  • Each NDB table must have a primary key. If no primary key is defined by the user when a table is created, the NDB Cluster storage engine automatically generates a hidden one. (Note: This hidden key takes up space just as does any other table index. It is not uncommon to encounter problems due to insufficient memory for accommodating these automatically created indexes.)

If you are importing tables from an existing database using the output of mysqldump, you can open the SQL script in a text editor and add the ENGINE option to any table creation statements, or replace any existing ENGINE (or TYPE) options. Suppose that you have the world sample database on another MySQL server that does not support MySQL Cluster, and you want to export the City table:

shell> mysqldump --add-drop-table world City > city_table.sql

The resulting city_table.sql file will contain this table creation statement (and the INSERT statements necessary to import the table data):

  `ID` int(11) NOT NULL auto_increment,
  `Name` char(35) NOT NULL default '',
  `CountryCode` char(3) NOT NULL default '',
  `District` char(20) NOT NULL default '',
  `Population` int(11) NOT NULL default '0',

INSERT INTO `City` VALUES (1,'Kabul','AFG','Kabol',1780000);
INSERT INTO `City` VALUES (2,'Qandahar','AFG','Qandahar',237500);
INSERT INTO `City` VALUES (3,'Herat','AFG','Herat',186800);
(remaining INSERT statements omitted)

You will need to make sure that MySQL uses the NDB storage engine for this table. There are two ways that this can be accomplished. One of these is to modify the table definition before importing it into the Cluster database. Using the City table as an example, modify the ENGINE option of the definition as follows:

  `ID` int(11) NOT NULL auto_increment,
  `Name` char(35) NOT NULL default '',
  `CountryCode` char(3) NOT NULL default '',
  `District` char(20) NOT NULL default '',
  `Population` int(11) NOT NULL default '0',

INSERT INTO `City` VALUES (1,'Kabul','AFG','Kabol',1780000);
INSERT INTO `City` VALUES (2,'Qandahar','AFG','Qandahar',237500);
INSERT INTO `City` VALUES (3,'Herat','AFG','Herat',186800);
(remaining INSERT statements omitted)

This must be done for the definition of each table that is to be part of the clustered database. The easiest way to accomplish this is to do a search-and-replace on the file that contains the definitions and replace all instances of TYPE=engine_name or ENGINE=engine_name with ENGINE=NDBCLUSTER. If you do not want to modify the file, you can use the unmodified file to create the tables, and then use ALTER TABLE to change their storage engine. The particulars are given later in this section.

Assuming that you have already created a database named world on the SQL node of the cluster, you can then use the mysql command-line client to read city_table.sql, and create and populate the corresponding table in the usual manner:

shell> mysql world < city_table.sql

It is very important to keep in mind that the preceding command must be executed on the host where the SQL node is running (in this case, on the machine with the IP address

To create a copy of the entire world database on the SQL node, use mysqldump on the non-cluster server to export the database to a file named world.sql; for example, in the /tmp directory. Then modify the table definitions as just described and import the file into the SQL node of the cluster like this:

shell> mysql world < /tmp/world.sql

If you save the file to a different location, adjust the preceding instructions accordingly.

It is important to note that NDB Cluster in MySQL 5.1 does not support autodiscovery of databases. (See Section 15.13, “Known Limitations of MySQL Cluster”.) This means that, once the world database and its tables have been created on one data node, you need to issue the CREATE SCHEMA world statement.

Running SELECT queries on the SQL node is no different from running them on any other instance of a MySQL server. To run queries from the command line, you first need to log in to the MySQL Monitor in the usual way (specify the root password at the Enter password: prompt):

shell> mysql -u root -p
Enter password:
Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 1 to server version: 5.1.15-beta

Type 'help;' or '\h' for help. Type '\c' to clear the buffer.


We simply use the MySQL server's root account and assume that you have followed the standard security precautions for installing a MySQL server, including setting a strong root password. For more information, see Section 2.10.3, “Securing the Initial MySQL Accounts”.

It is worth taking into account that Cluster nodes do not make use of the MySQL privilege system when accessing one another. Setting or changing MySQL user accounts (including the root account) effects only applications that access the SQL node, not interaction between nodes.

If you did not modify the ENGINE clauses in the table definitions prior to importing the SQL script, you should run the following statements at this point:

mysql> USE world;

Selecting a database and running a SELECT query against a table in that database is also accomplished in the usual manner, as is exiting the MySQL Monitor:

mysql> USE world;
mysql> SELECT Name, Population FROM City ORDER BY Population DESC LIMIT 5;
| Name      | Population |
| Bombay    |   10500000 |
| Seoul     |    9981619 |
| São Paulo |    9968485 |
| Shanghai  |    9696300 |
| Jakarta   |    9604900 |
5 rows in set (0.34 sec)

mysql> \q


Applications that use MySQL can employ standard APIs to access NDB tables. It is important to remember that your application must access the SQL node, and not the MGM or data nodes. This brief example shows how we might execute the SELECT statement just shown by using PHP 5's mysqli extension running on a Web server elsewhere on the network:

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
  <meta http-equiv="Content-Type"
        content="text/html; charset=iso-8859-1">
  <title>SIMPLE mysqli SELECT</title>
  # connect to SQL node:
  $link = new mysqli('', 'root', 'root_password', 'world');
  # parameters for mysqli constructor are:
  #   host, user, password, database

  if( mysqli_connect_errno() )
    die("Connect failed: " . mysqli_connect_error());

  $query = "SELECT Name, Population
            FROM City
            ORDER BY Population DESC
            LIMIT 5";

  # if no errors...
  if( $result = $link->query($query) )
<table border="1" width="40%" cellpadding="4" cellspacing ="1">
    <th width="10%">City</th>
    # then display the results...
    while($row = $result->fetch_object())
      printf(<tr>\n  <td align=\"center\">%s</td><td>%d</td>\n</tr>\n",
              $row->Name, $row->Population);
  # ...and verify the number of rows that were retrieved
    printf("<p>Affected rows: %d</p>\n", $link->affected_rows);
    # otherwise, tell us what went wrong
    echo mysqli_error();

  # free the result set and the mysqli connection object

We assume that the process running on the Web server can reach the IP address of the SQL node.

In a similar fashion, you can use the MySQL C API, Perl-DBI, Python-mysql, or MySQL AB's own Connectors to perform the tasks of data definition and manipulation just as you would normally with MySQL.

15.3.6. Safe Shutdown and Restart

To shut down the cluster, enter the following command in a shell on the machine hosting the MGM node:

shell> ndb_mgm -e shutdown

The -e option here is used to pass a command to the ndb_mgm client from the shell. See Section 4.3.1, “Using Options on the Command Line”. The command causes the ndb_mgm, ndb_mgmd, and any ndbd processes to terminate gracefully. Any SQL nodes can be terminated using mysqladmin shutdown and other means.

To restart the cluster, run these commands:

  • On the management host ( in our example setup):

    shell> ndb_mgmd -f /var/lib/mysql-cluster/config.ini
  • On each of the data node hosts ( and

    shell> ndbd

    Remember not to invoke this command with the --initial option when restarting an NDBD node normally.

  • On the SQL host (

    shell> mysqld &

For information on making Cluster backups, see Section 15.8.2, “Using The Management Client to Create a Backup”.

To restore the cluster from backup requires the use of the ndb_restore command. This is covered in Section 15.8.3, “How to Restore a Cluster Backup”.

More information on configuring MySQL Cluster can be found in Section 15.4, “MySQL Cluster Configuration”.

15.4. MySQL Cluster Configuration

A MySQL server that is part of a MySQL Cluster differs in only one respect from a normal (non-clustered) MySQL server, in that it employs the NDB Cluster storage engine. This engine is also referred to simply as NDB, and the two forms of the name are synonymous.

To avoid unnecessary allocation of resources, the server is configured by default with the NDB storage engine disabled. To enable NDB, you must modify the server's my.cnf configuration file, or start the server with the --ndbcluster option.

The MySQL server is a part of the cluster, so it also must know how to access an MGM node to obtain the cluster configuration data. The default behavior is to look for the MGM node on localhost. However, should you need to specify that its location is elsewhere, this can be done in my.cnf or on the MySQL server command line. Before the NDB storage engine can be used, at least one MGM node must be operational, as well as any desired data nodes.

15.4.1. Building MySQL Cluster from Source Code

NDB, the Cluster storage engine, is available in binary distributions for Linux, Mac OS X, and Solaris. We are working to make Cluster run on all operating systems supported by MySQL, including Windows.

If you choose to build from a source tarball or the MySQL 5.1 BitKeeper tree, be sure to use the --with-ndbcluster option when running configure. You can also use the BUILD/compile-pentium-max build script. Note that this script includes OpenSSL, so you must either have or obtain OpenSSL to build successfully, or else modify compile-pentium-max to exclude this requirement. Of course, you can also just follow the standard instructions for compiling your own binaries, and then perform the usual tests and installation procedure. See Section 2.9.3, “Installing from the Development Source Tree”.

15.4.2. Installing the Software

In the next few sections, we assume that you are already familiar with installing MySQL, and here we cover only the differences between configuring MySQL Cluster and configuring MySQL without clustering. (See Chapter 2, Installing and Upgrading MySQL, if you require more information about the latter.)

You will find Cluster configuration easiest if you have already have all management and data nodes running first; this is likely to be the most time-consuming part of the configuration. Editing the my.cnf file is fairly straightforward, and this section will cover only any differences from configuring MySQL without clustering.

15.4.3. Quick Test Setup of MySQL Cluster

To familiarize you with the basics, we will describe the simplest possible configuration for a functional MySQL Cluster. After this, you should be able to design your desired setup from the information provided in the other relevant sections of this chapter.

First, you need to create a configuration directory such as /var/lib/mysql-cluster, by executing the following command as the system root user:

shell> mkdir /var/lib/mysql-cluster

In this directory, create a file named config.ini that contains the following information. Substitute appropriate values for HostName and DataDir as necessary for your system.

# file "config.ini" - showing minimal setup consisting of 1 data node,
# 1 management server, and 3 MySQL servers.
# The empty default sections are not required, and are shown only for
# the sake of completeness.
# Data nodes must provide a hostname but MySQL Servers are not required
# to do so.
# If you don't know the hostname for your machine, use localhost.
# The DataDir parameter also has a default value, but it is recommended to
# set it explicitly.
# Note: DB, API, and MGM are aliases for NDBD, MYSQLD, and NDB_MGMD
# respectively. DB and API are deprecated and should not be used in new
# installations.
NoOfReplicas= 1



DataDir= /var/lib/mysql-cluster


You can now start the ndb_mgmd management server. By default, it atttempts to read the config.ini file in its current working directory, so change location into the directory where the file is located and then invoke ndb_mgmd:

shell> cd /var/lib/mysql-cluster
shell> ndb_mgmd

Then start a single data node by running ndbd. When starting ndbd for a given data node for the very first time, you should use the --initial option as shown here:

shell> ndbd --initial

For subsequent ndbd starts, you will generally want to omit the --initial option:

shell> ndbd

The reason for omitting --initial on subsequent restarts is that this option causes ndbd to delete and re-create all existing data and log files (as well as all table metadata) for this data node. One exception to this rule about not using --initial except for the first ndbd invocation is that you use it when restarting the cluster and restoring from backup after adding new data nodes.

By default, ndbd looks for the management server at localhost on port 1186.

Note: If you have installed MySQL from a binary tarball, you will need to specify the path of the ndb_mgmd and ndbd servers explicitly. (Normally, these will be found in /usr/local/mysql/bin.)

Finally, change location to the MySQL data directory (usually /var/lib/mysql or /usr/local/mysql/data), and make sure that the my.cnf file contains the option necessary to enable the NDB storage engine:


You can now start the MySQL server as usual:

shell> mysqld_safe --user=mysql &

Wait a moment to make sure the MySQL server is running properly. If you see the notice mysql ended, check the server's .err file to find out what went wrong.

If all has gone well so far, you now can start using the cluster. Connect to the server and verify that the NDBCLUSTER storage engine is enabled:

shell> mysql
Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 1 to server version: 5.1.15-beta

Type 'help;' or '\h' for help. Type '\c' to clear the buffer.

*************************** 12. row ***************************
Support: YES
Comment: Clustered, fault-tolerant, memory-based tables
*************************** 13. row ***************************
Engine: NDB
Support: YES
Comment: Alias for NDBCLUSTER

The row numbers shown in the preceding example output may be different from those shown on your system, depending upon how your server is configured.

Try to create an NDBCLUSTER table:

shell> mysql
mysql> USE test;
Database changed

Query OK, 0 rows affected (0.09 sec)

mysql> SHOW CREATE TABLE ctest \G
*************************** 1. row ***************************
       Table: ctest
Create Table: CREATE TABLE `ctest` (
  `i` int(11) default NULL
) ENGINE=ndbcluster DEFAULT CHARSET=latin1
1 row in set (0.00 sec)

To check that your nodes were set up properly, start the management client:

shell> ndb_mgm

Use the SHOW command from within the management client to obtain a report on the cluster's status:

Cluster Configuration
[ndbd(NDB)]     1 node(s)
id=2    @  (Version: 3.5.3, Nodegroup: 0, Master)

[ndb_mgmd(MGM)] 1 node(s)
id=1    @  (Version: 3.5.3)

[mysqld(API)]   3 node(s)
id=3    @  (Version: 3.5.3)
id=4 (not connected, accepting connect from any host)
id=5 (not connected, accepting connect from any host)

At this point, you have successfully set up a working MySQL Cluster. You can now store data in the cluster by using any table created with ENGINE=NDBCLUSTER or its alias ENGINE=NDB.

15.4.4. Configuration File

Configuring MySQL Cluster requires working with two files:

  • my.cnf: Specifies options for all MySQL Cluster executables. This file, with which you should be familiar with from previous work with MySQL, must be accessible by each executable running in the cluster.

  • config.ini: This file is read only by the MySQL Cluster management server, which then distributes the information contained therein to all processes participating in the cluster. config.ini contains a description of each node involved in the cluster. This includes configuration parameters for data nodes and configuration parameters for connections between all nodes in the cluster. For a quick reference to the sections that can appear in this file, and what sorts of configuration parameters may be placed in each section, see Sections of the config.ini File.

We are continuously making improvements in Cluster configuration and attempting to simplify this process. Although we strive to maintain backward compatibility, there may be times when introduce an incompatible change. In such cases we will try to let Cluster users know in advance if a change is not backward compatible. If you find such a change and we have not documented it, please report it in the MySQL bugs database using the instructions given in Section 1.8, “How to Report Bugs or Problems”. Basic Example Configuration

To support MySQL Cluster, you will need to update my.cnf as shown in the following example. Note that the options shown here should not be confused with those that are used in config.ini files. You may also specify these parameters on the command line when invoking the executables.

# my.cnf
# example additions to my.cnf for MySQL Cluster
# (valid in MySQL 5.1)

# enable ndbcluster storage engine, and provide connectstring for
# management server host (default port is 1186)

# provide connectstring for management server host (default port: 1186)

# provide connectstring for management server host (default port: 1186)

# provide location of cluster configuration file

(For more information on connectstrings, see Section, “The Cluster connectstring.)

# my.cnf
# example additions to my.cnf for MySQL Cluster
# (will work on all versions)

# enable ndbcluster storage engine, and provide connectstring for management
# server host to the default port 1186

Important: Once you have started a mysqld process with the ndbcluster and ndb-connectstring parameters in the [MYSQLD] in the my.cnf file as shown previously, you cannot execute any CREATE TABLE or ALTER TABLE statements without having actually started the cluster. Otherwise, these statements will fail with an error. This is by design.

You may also use a separate [mysql_cluster] section in the cluster my.cnf file for settings to be read and used by all executables:

# cluster-specific settings

For additional NDB variables that can be set in the my.cnf file, see Section 5.2.3, “System Variables”.

The configuration file is named config.ini by default. It is read by ndb_mgmd at startup and can be placed anywhere. Its location and name are specified by using --config-file=path_name on the ndb_mgmd command line. If the configuration file is not specified, ndb_mgmd by default tries to read a file named config.ini located in the current working directory.

Currently, the configuration file is in INI format, which consists of sections preceded by section headings (surrounded by square brackets), followed by the appropriate parameter names and values. One deviation from the standard INI format is that the parameter name and value can be separated by a colon (‘:’) as well as the equals sign (‘=’). Another deviation is that sections are not uniquely identified by section name. Instead, unique sections (such as two different nodes of the same type) are identified by a unique ID specified as a parameter within the section.

Default values are defined for most parameters, and can also be specified in config.ini. To create a default value section, simply add the word DEFAULT to the section name. For example, an [NDBD] section contains parameters that apply to a particular data node, whereas an [NDBD DEFAULT] section contains parameters that apply to all data nodes. Suppose that all data nodes should use the same data memory size. To configure them all, create an [NDBD DEFAULT] section that contains a DataMemory line to specify the data memory size.

At a minimum, the configuration file must define the computers and nodes involved in the cluster and on which computers these nodes are located. An example of a simple configuration file for a cluster consisting of one management server, two data nodes and two MySQL servers is shown here:

# file "config.ini" - 2 data nodes and 2 SQL nodes
# This file is placed in the startup directory of ndb_mgmd (the
# management server)
# The first MySQL Server can be started from any host. The second
# can be started only on the host

NoOfReplicas= 2
DataDir= /var/lib/mysql-cluster

DataDir= /var/lib/mysql-cluster




Note that each node has its own section in the config.ini. For instance, this cluster has two data nodes, so the preceding configuration file contains two [NDBD] sections defining these nodes.

Sections of the config.ini File

There are six different sections that you can use in the config.ini configuration file, as described in the following list:

You can define DEFAULT values for each section. All Cluster parameter names are case-insensitive, which differs from parameters specified in my.cnf or my.ini files. The Cluster connectstring

With the exception of the MySQL Cluster management server (ndb_mgmd), each node that is part of a MySQL Cluster requires a connectstring that points to the management server's location. This connectstring is used in establishing a connection to the management server as well as in performing other tasks depending on the node's role in the cluster. The syntax for a connectstring is as follows:

<connectstring> :=

<nodeid-specification> := node_id

<host-specification> := host_name[:port_num]

node_id is an integer larger than 1 which identifies a node in config.ini. host_name is a string representing a valid Internet host name or IP address. port_num is an integer referring to a TCP/IP port number.

example 1 (long):    "nodeid=2,myhost1:1100,myhost2:1100,"
example 2 (short):   "myhost1"

All nodes will use localhost:1186 as the default connectstring value if none is provided. If port_num is omitted from the connectstring, the default port is 1186. This port should always be available on the network because it has been assigned by IANA for this purpose (see for details).

By listing multiple <host-specification> values, it is possible to designate several redundant management servers. A cluster node will attempt to contact successive management servers on each host in the order specified, until a successful connection has been established.

There are a number of different ways to specify the connectstring:

  • Each executable has its own command-line option which enables specifying the management server at startup. (See the documentation for the respective executable.)

  • It is also possible to set the connectstring for all nodes in the cluster at once by placing it in a [mysql_cluster] section in the management server's my.cnf file.

  • For backward compatibility, two other options are available, using the same syntax:

    1. Set the NDB_CONNECTSTRING environment variable to contain the connectstring.

    2. Write the connectstring for each executable into a text file named Ndb.cfg and place this file in the executable's startup directory.

    However, these are now deprecated and should not be used for new installations.

The recommended method for specifying the connectstring is to set it on the command line or in the my.cnf file for each executable. Defining Cluster Computers

The [COMPUTER] section has no real significance other than serving as a way to avoid the need of defining host names for each node in the system. All parameters mentioned here are required.

  • Id

    This is an integer value, used to refer to the host computer elsewhere in the configuration file. This is not the same as the node ID.

  • HostName

    This is the computer's hostname or IP address. Defining the Management Server

The [NDB_MGMD] section is used to configure the behavior of the management server. [MGM] can be used as an alias; the two section names are equivalent. All parameters in the following list are optional and assume their default values if omitted. Note: If neither the ExecuteOnComputer nor the HostName parameter is present, the default value localhost will be assumed for both.

  • Id

    Each node in the cluster has a unique identity, which is represented by an integer value in the range 1 to 63 inclusive. This ID is used by all internal cluster messages for addressing the node.

  • ExecuteOnComputer

    This refers to the Id set for one of the computers defined in a [COMPUTER] section of the config.ini file.

  • PortNumber

    This is the port number on which the management server listens for configuration requests and management commands.

  • HostName

    Specifying this parameter defines the hostname of the computer on which the management node is to reside. To specify a hostname other than localhost, either this parameter or ExecuteOnComputer is required.

  • LogDestination

    This parameter specifies where to send cluster logging information. There are three options in this regard — CONSOLE, SYSLOG, and FILE — with FILE being the default:

    • CONSOLE outputs the log to stdout:

    • SYSLOG sends the log to a syslog facility, possible values being one of auth, authpriv, cron, daemon, ftp, kern, lpr, mail, news, syslog, user, uucp, local0, local1, local2, local3, local4, local5, local6, or local7.

      Note: Not every facility is necessarily supported by every operating system.

    • FILE pipes the cluster log output to a regular file on the same machine. The following values can be specified:

      • filename: The name of the log file.

      • maxsize: The maximum size (in bytes) to which the file can grow before logging rolls over to a new file. When this occurs, the old log file is renamed by appending .N to the filename, where N is the next number not yet used with this name.

      • maxfiles: The maximum number of log files.


      The default value for the FILE parameter is FILE:filename=ndb_node_id_cluster.log,maxsize=1000000,maxfiles=6, where node_id is the ID of the node.

    It is possible to specify multiple log destinations separated by semicolons as shown here:

  • ArbitrationRank

    This parameter is used to define which nodes can act as arbitrators. Only MGM nodes and SQL nodes can be arbitrators. ArbitrationRank can take one of the following values:

    • 0: The node will never be used as an arbitrator.

    • 1: The node has high priority; that is, it will be preferred as an arbitrator over low-priority nodes.

    • 2: Indicates a low-priority node which be used as an arbitrator only if a node with a higher priority is not available for that purpose.

    Normally, the management server should be configured as an arbitrator by setting its ArbitrationRank to 1 (the default value) and that of all SQL nodes to 0.

  • ArbitrationDelay

    An integer value which causes the management server's responses to arbitration requests to be delayed by that number of milliseconds. By default, this value is 0; it is normally not necessary to change it.

  • DataDir

    This specifies the directory where output files from the management server will be placed. These files include cluster log files, process output files, and the daemon's process ID (PID) file. (For log files, this location can be overridden by setting the FILE parameter for LogDestination as discussed previously in this section.)

    The default value for this parameter is the directory in which ndb_mgmd is located. Defining Data Nodes

The [NDBD] and [NDBD DEFAULT] sections are used to configure the behavior of the cluster's data nodes. There are many parameters which control buffer sizes, pool sizes, timeouts, and so forth. The only mandatory parameters are:

  • Either ExecuteOnComputer or HostName, which must be defined in the local [NDBD] section.

  • The parameter NoOfReplicas, which must be defined in the [NDBD DEFAULT] section, as it is common to all Cluster data nodes.

Most data node parameters are set in the [NDBD DEFAULT] section. Only those parameters explicitly stated as being able to set local values are allowed to be changed in the [NDBD] section. Where present, HostName, Id and ExecuteOnComputer must be defined in the local [NDBD] section, and not in any other section of config.ini. In other words, settings for these parameters are specific to one data node.

For those parameters affecting memory usage or buffer sizes, it is possible to use K, M, or G as a suffix to indicate units of 1024, 1024×1024, or 1024×1024×1024. (For example, 100K means 100 × 1024 = 102400.) Parameter names and values are currently case-sensitive.

Identifying Data Nodes

The Id value (that is, the data node identifier) can be allocated on the command line when the node is started or in the configuration file.

  • Id

    This is the node ID used as the address of the node for all cluster internal messages. This is an integer in the range 1 to 63 inclusive. Each node in the cluster must have a unique identity.

  • ExecuteOnComputer

    This refers to the Id set for one of the computers defined in a [COMPUTER] section.

  • HostName

    Specifying this parameter defines the hostname of the computer on which the data node is to reside. To specify a hostname other than localhost, either this parameter or ExecuteOnComputer is required.

  • ServerPort (OBSOLETE)

    Each node in the cluster uses a port to connect to other nodes. This port is used also for non-TCP transporters in the connection setup phase. The default port is allocated dynamically in such a way as to ensure that no two nodes on the same computer receive the same port number, so it should not normally be necessary to specify a value for this parameter.

  • NoOfReplicas

    This global parameter can be set only in the [NDBD DEFAULT] section, and defines the number of replicas for each table stored in the cluster. This parameter also specifies the size of node groups. A node group is a set of nodes all storing the same information.

    Node groups are formed implicitly. The first node group is formed by the set of data nodes with the lowest node IDs, the next node group by the set of the next lowest node identities, and so on. By way of example, assume that we have 4 data nodes and that NoOfReplicas is set to 2. The four data nodes have node IDs 2, 3, 4 and 5. Then the first node group is formed from nodes 2 and 3, and the second node group by nodes 4 and 5. It is important to configure the cluster in such a manner that nodes in the same node groups are not placed on the same computer because a single hardware failure would cause the entire cluster to crash.

    If no node IDs are provided, the order of the data nodes will be the determining factor for the node group. Whether or not explicit assignments are made, they can be viewed in the output of the management client's SHOW statement.

    There is no default value for NoOfReplicas; the maximum possible value is 4.

    Important: The value for this parameter must divide evenly into the number of data nodes in the cluster. For example, if there are two data nodes, then NoOfReplicas must be equal to either 1 or 2, since 2/3 and 2/4 both yield fractional values; if there are four data nodes, then NoOfReplicas must be equal to 1, 2, or 4.

  • DataDir

    This parameter specifies the directory where trace files, log files, pid files and error logs are placed.

  • FileSystemPath

    This parameter specifies the directory where all files created for metadata, REDO logs, UNDO logs and data files are placed. The default is the directory specified by DataDir. Note: This directory must exist before the ndbd process is initiated.

    The recommended directory hierarchy for MySQL Cluster includes /var/lib/mysql-cluster, under which a directory for the node's filesystem is created. The name of this subdirectory contains the node ID. For example, if the node ID is 2, this subdirectory is named ndb_2_fs.

  • BackupDataDir

    This parameter specifies the directory in which backups are placed. If omitted, the default backup location is the directory named BACKUP under the location specified by the FileSystemPath parameter. (See above.)

Data Memory, Index Memory, and String Memory

DataMemory and IndexMemory are [NDBD] parameters specifying the size of memory segments used to store the actual records and their indexes. In setting values for these, it is important to understand how DataMemory and IndexMemory are used, as they usually need to be updated to reflect actual usage by the cluster:

  • DataMemory

    This parameter defines the amount of space (in bytes) available for storing database records. The entire amount specified by this value is allocated in memory, so it is extremely important that the machine has sufficient physical memory to accommodate it.

    The memory allocated by DataMemory is used to store both the actual records and indexes. There is a 16-byte overhead on each record; an additional amount for each record is incurred because it is stored in a 32KB page with 128 byte page overhead (see below). There is also a small amount wasted per page due to the fact that each record is stored in only one page.

    For variable-size table attributes in MySQL 5.1, the data is stored on separate datapages, allocated from DataMemory. Variable-length records use a fixed-size part with an extra overhead of 4 bytes to reference the variable-size part. The variable-size part has 2 bytes overhead plus 2 bytes per attribute.

    The maximum record size is currently 8052 bytes.

    The memory space defined by DataMemory is also used to store ordered indexes, which use about 10 bytes per record. Each table row is represented in the ordered index. A common error among users is to assume that all indexes are stored in the memory allocated by IndexMemory, but this is not the case: Only primary key and unique hash indexes use this memory; ordered indexes use the memory allocated by DataMemory. However, creating a primary key or unique hash index also creates an ordered index on the same keys, unless you specify USING HASH in the index creation statement. This can be verified by running ndb_desc -d db_name table_name in the management client.

    The memory space allocated by DataMemory consists of 32KB pages, which are allocated to table fragments. Each table is normally partitioned into the same number of fragments as there are data nodes in the cluster. Thus, for each node, there are the same number of fragments as are set in NoOfReplicas.

    Once a page has been allocated, it is currently not possible to return it to the pool of free pages, except by deleting the table. (This also means that DataMemory pages, once allocated to a given table, cannot be used by other tables.) Performing a node recovery also compresses the partition because all records are inserted into empty partitions from other live nodes.

    The DataMemory memory space also contains UNDO information: For each update, a copy of the unaltered record is allocated in the DataMemory. There is also a reference to each copy in the ordered table indexes. Unique hash indexes are updated only when the unique index columns are updated, in which case a new entry in the index table is inserted and the old entry is deleted upon commit. For this reason, it is also necessary to allocate enough memory to handle the largest transactions performed by applications using the cluster. In any case, performing a few large transactions holds no advantage over using many smaller ones, for the following reasons:

    • Large transactions are not any faster than smaller ones

    • Large transactions increase the number of operations that are lost and must be repeated in event of transaction failure

    • Large transactions use more memory

    The default value for DataMemory is 80MB; the minimum is 1MB. There is no maximum size, but in reality the maximum size has to be adapted so that the process does not start swapping when the limit is reached. This limit is determined by the amount of physical RAM available on the machine and by the amount of memory that the operating system may commit to any one process. 32-bit operating systems are generally limited to 2–4GB per process; 64-bit operating systems can use more. For large databases, it may be preferable to use a 64-bit operating system for this reason.

  • IndexMemory

    This parameter controls the amount of storage used for hash indexes in MySQL Cluster. Hash indexes are always used for primary key indexes, unique indexes, and unique constraints. Note that when defining a primary key and a unique index, two indexes will be created, one of which is a hash index used for all tuple accesses as well as lock handling. It is also used to enforce unique constraints.

    The size of the hash index is 25 bytes per record, plus the size of the primary key. For primary keys larger than 32 bytes another 8 bytes is added.

    The default value for IndexMemory is 18MB. The minimum is 1MB.

  • StringMemory

    This parameter determines how much memory is allocated for strings such as table names, and is specified in an [NDBD] or [NDBD DEFAULT] section of the config.ini file. A value between 0 and 100 inclusive is interpreted as a percent of the maxmimum default value, which is calculated based on a number of factors including the number of tables, maximum table name size, maximum size of .FRM files, MaxNoOfTriggers, maximum column name size, and maximum default column value. In general it is safe to assume that the maximum default value is approximately 5 MB for a MySQL Cluster having 1000 tables.

    A value greater than 100 is interpreted as a number of bytes.

    The default value is 5 — that is, 5 percent of the default maximum, or roughly 5 KB. (Note that this is a change from previous versions of MySQL Cluster.)

    Under most circumstances, the default value should be sufficient, but when you have a great many Cluster tables (1000 or more), it is possible to get Error 773 Out of string memory, please modify StringMemory config parameter: Permanent error: Schema error, in which case you should increase this value. 25 (25 percent) is not excessive, and should prevent this error from recurring in all but the most extreme conditions.

The following example illustrates how memory is used for a table. Consider this table definition:

CREATE TABLE example (

For each record, there are 12 bytes of data plus 12 bytes overhead. Having no nullable columns saves 4 bytes of overhead. In addition, we have two ordered indexes on columns a and b consuming roughly 10 bytes each per record. There is a primary key hash index on the base table using roughly 29 bytes per record. The unique constraint is implemented by a separate table with b as primary key and a as a column. This other table consumes an additional 29 bytes of index memory per record in the example table as well 8 bytes of record data plus 12 bytes of overhead.

Thus, for one million records, we need 58MB for index memory to handle the hash indexes for the primary key and the unique constraint. We also need 64MB for the records of the base table and the unique index table, plus the two ordered index tables.

You can see that hash indexes takes up a fair amount of memory space; however, they provide very fast access to the data in return. They are also used in MySQL Cluster to handle uniqueness constraints.

Currently, the only partitioning algorithm is hashing and ordered indexes are local to each node. Thus, ordered indexes cannot be used to handle uniqueness constraints in the general case.

An important point for both IndexMemory and DataMemory is that the total database size is the sum of all data memory and all index memory for each node group. Each node group is used to store replicated information, so if there are four nodes with two replicas, there will be two node groups. Thus, the total data memory available is 2 × DataMemory for each data node.

It is highly recommended that DataMemory and IndexMemory be set to the same values for all nodes. Data distribution is even over all nodes in the cluster, so the maximum amount of space available for any node can be no greater than that of the smallest node in the cluster.

DataMemory and IndexMemory can be changed, but decreasing either of these can be risky; doing so can easily lead to a node or even an entire MySQL Cluster that is unable to restart due to there being insufficient memory space. Increasing these values should be acceptable, but it is recommended that such upgrades are performed in the same manner as a software upgrade, beginning with an update of the configuration file, and then restarting the management server followed by restarting each data node in turn.

Updates do not increase the amount of index memory used. Inserts take effect immediately; however, rows are not actually deleted until the transaction is committed.

Transaction Parameters

The next three [NDBD] parameters that we discuss are important because they affect the number of parallel transactions and the sizes of transactions that can be handled by the system. MaxNoOfConcurrentTransactions sets the number of parallel transactions possible in a node. MaxNoOfConcurrentOperations sets the number of records that can be in update phase or locked simultaneously.

Both of these parameters (especially MaxNoOfConcurrentOperations) are likely targets for users setting specific values and not using the default value. The default value is set for systems using small transactions, to ensure that these do not use excessive memory.

  • MaxNoOfConcurrentTransactions

    For each active transaction in the cluster there must be a record in one of the cluster nodes. The task of coordinating transactions is spread among the nodes. The total number of transaction records in the cluster is the number of transactions in any given node times the number of nodes in the cluster.

    Transaction records are allocated to individual MySQL servers. Normally, there is at least one transaction record allocated per connection that using any table in the cluster. For this reason, one should ensure that there are more transaction records in the cluster than there are concurrent connections to all MySQL servers in the cluster.

    This parameter must be set to the same value for all cluster nodes.

    Changing this parameter is never safe and doing so can cause a cluster to crash. When a node crashes, one of the nodes (actually the oldest surviving node) will build up the transaction state of all transactions ongoing in the crashed node at the time of the crash. It is thus important that this node has as many transaction records as the failed node.

    The default value is 4096.

  • MaxNoOfConcurrentOperations

    It is a good idea to adjust the value of this parameter according to the size and number of transactions. When performing transactions of only a few operations each and not involving a great many records, there is no need to set this parameter very high. When performing large transactions involving many records need to set this parameter higher.

    Records are kept for each transaction updating cluster data, both in the transaction coordinator and in the nodes where the actual updates are performed. These records contain state information needed to find UNDO records for rollback, lock queues, and other purposes.

    This parameter should be set to the number of records to be updated simultaneously in transactions, divided by the number of cluster data nodes. For example, in a cluster which has four data nodes and which is expected to handle 1,000,000 concurrent updates using transactions, you should set this value to 1000000 / 4 = 250000.

    Read queries which set locks also cause operation records to be created. Some extra space is allocated within individual nodes to accommodate cases where the distribution is not perfect over the nodes.

    When queries make use of the unique hash index, there are actually two operation records used per record in the transaction. The first record represents the read in the index table and the second handles the operation on the base table.

    The default value is 32768.

    This parameter actually handles two values that can be configured separately. The first of these specifies how many operation records are to be placed with the transaction coordinator. The second part specifies how many operation records are to be local to the database.

    A very large transaction performed on an eight-node cluster requires as many operation records in the transaction coordinator as there are reads, updates, and deletes involved in the transaction. However, the operation records of the are spread over all eight nodes. Thus, if it is necessary to configure the system for one very large transaction, it is a good idea to configure the two parts separately. MaxNoOfConcurrentOperations will always be used to calculate the number of operation records in the transaction coordinator portion of the node.

    It is also important to have an idea of the memory requirements for operation records. These consume about 1KB per record.

  • MaxNoOfLocalOperations

    By default, this parameter is calculated as 1.1 × MaxNoOfConcurrentOperations. This fits systems with many simultaneous transactions, none of them being very large. If there is a need to handle one very large transaction at a time and there are many nodes, it is a good idea to override the default value by explicitly specifying this parameter.

Transaction Temporary Storage

The next set of [NDBD] parameters is used to determine temporary storage when executing a statement that is part of a Cluster transaction. All records are released when the statement is completed and the cluster is waiting for the commit or rollback.

The default values for these parameters are adequate for most situations. However, users with a need to support transactions involving large numbers of rows or operations may need to increase these values to enable better parallelism in the system, whereas users whose applications require relatively small transactions can decrease the values to save memory.

  • MaxNoOfConcurrentIndexOperations

    For queries using a unique hash index, another temporary set of operation records is used during a query's execution phase. This parameter sets the size of that pool of records. Thus, this record is allocated only while executing a part of a query. As soon as this part has been executed, the record is released. The state needed to handle aborts and commits is handled by the normal operation records, where the pool size is set by the parameter MaxNoOfConcurrentOperations.

    The default value of this parameter is 8192. Only in rare cases of extremely high parallelism using unique hash indexes should it be necessary to increase this value. Using a smaller value is possible and can save memory if the DBA is certain that a high degree of parallelism is not required for the cluster.

  • MaxNoOfFiredTriggers

    The default value of MaxNoOfFiredTriggers is 4000, which is sufficient for most situations. In some cases it can even be decreased if the DBA feels certain the need for parallelism in the cluster is not high.

    A record is created when an operation is performed that affects a unique hash index. Inserting or deleting a record in a table with unique hash indexes or updating a column that is part of a unique hash index fires an insert or a delete in the index table. The resulting record is used to represent this index table operation while waiting for the original operation that fired it to complete. This operation is short-lived but can still require a large number of records in its pool for situations with many parallel write operations on a base table containing a set of unique hash indexes.

  • TransactionBufferMemory

    The memory affected by this parameter is used for tracking operations fired when updating index tables and reading unique indexes. This memory is used to store the key and column information for these operations. It is only very rarely that the value for this parameter needs to be altered from the default.

    The default value for TransactionBufferMemory is 1MB.

    Normal read and write operations use a similar buffer, whose usage is even more short-lived. The compile-time parameter ZATTRBUF_FILESIZE (found in ndb/src/kernel/blocks/Dbtc/Dbtc.hpp) set to 4000 × 128 bytes (500KB). A similar buffer for key information, ZDATABUF_FILESIZE (also in Dbtc.hpp) contains 4000 × 16 = 62.5KB of buffer space. Dbtc is the module that handles transaction coordination.

Scans and Buffering

There are additional [NDBD] parameters in the Dblqh module (in ndb/src/kernel/blocks/Dblqh/Dblqh.hpp) that affect reads and updates. These include ZATTRINBUF_FILESIZE, set by default to 10000 × 128 bytes (1250KB) and ZDATABUF_FILE_SIZE, set by default to 10000*16 bytes (roughly 156KB) of buffer space. To date, there have been neither any reports from users nor any results from our own extensive tests suggesting that either of these compile-time limits should be increased.

  • MaxNoOfConcurrentScans

    This parameter is used to control the number of parallel scans that can be performed in the cluster. Each transaction coordinator can handle the number of parallel scans defined for this parameter. Each scan query is performed by scanning all partitions in parallel. Each partition scan uses a scan record in the node where the partition is located, the number of records being the value of this parameter times the number of nodes. The cluster should be able to sustain MaxNoOfConcurrentScans scans concurrently from all nodes in the cluster.

    Scans are actually performed in two cases. The first of these cases occurs when no hash or ordered indexes exists to handle the query, in which case the query is executed by performing a full table scan. The second case is encountered when there is no hash index to support the query but there is an ordered index. Using the ordered index means executing a parallel range scan. The order is kept on the local partitions only, so it is necessary to perform the index scan on all partitions.

    The default value of MaxNoOfConcurrentScans is 256. The maximum value is 500.

    This parameter specifies the number of scans possible in the transaction coordinator. If the number of local scan records is not provided, it is calculated as the product of MaxNoOfConcurrentScans and the number of data nodes in the system.

  • MaxNoOfLocalScans

    Specifies the number of local scan records if many scans are not fully parallelized.

  • BatchSizePerLocalScan

    This parameter is used to calculate the number of lock records which must be there to handle many concurrent scan operations.

    The default value is 64; this value has a strong connection to the ScanBatchSize defined in the SQL nodes.

  • LongMessageBuffer

    This is an internal buffer used for passing messages within individual nodes and between nodes. Although it is highly unlikely that this would need to be changed, it is configurable. By default, it is set to 1MB.

Logging and Checkpointing

These [NDBD] parameters control log and checkpoint behavior.

  • NoOfFragmentLogFiles

    This parameter sets the number of REDO log files for the node, and thus the amount of space allocated to REDO logging. Because the REDO log files are organized in a ring, it is extremely important that the first and last log files in the set (sometimes referred to as the “head” and “tail” log files, respectively) do not meet. When these approach one another too closely, the node begins aborting all transactions encompassing updates due to a lack of room for new log records.

    A REDO log record is not removed until three local checkpoints have been completed since that log record was inserted. Checkpointing frequency is determined by its own set of configuration parameters discussed elsewhere in this chapter.

    How these parameters interact and proposals for how to configure them are discussed in Section 15.4.6, “Configuring Parameters for Local Checkpoints”.

    The default parameter value is 8, which means 8 sets of 4 16MB files for a total of 512MB. In other words, REDO log space must be allocated in blocks of 64MB. In scenarios requiring a great many updates, the value for NoOfFragmentLogFiles may need to be set as high as 300 or even higher to provide sufficient space for REDO logs.

    If the checkpointing is slow and there are so many writes to the database that the log files are full and the log tail cannot be cut without jeopardizing recovery, all updating transactions are aborted with internal error code 410 (Out of log file space temporarily). This condition prevails until a checkpoint has completed and the log tail can be moved forward.

    Important: This parameter cannot be changed “on the fly”; you must restart the node using --initial. If you wish to change this value for a running cluster, you can do so via a rolling node restart.

  • MaxNoOfOpenFiles

    This parameter sets a ceiling on how many internal threads to allocate for open files. Any situation requiring a change in this parameter should be reported as a bug.

    The default value is 40.

  • MaxNoOfSavedMessages

    This parameter sets the maximum number of trace files that are kept before overwriting old ones. Trace files are generated when, for whatever reason, the node crashes.

    The default is 25 trace files.

Metadata Objects

The next set of [NDBD] parameters defines pool sizes for metadata objects, used to define the maximum number of attributes, tables, indexes, and trigger objects used by indexes, events, and replication between clusters. Note that these act merely as “suggestions” to the cluster, and any that are not specified revert to the default values shown.

  • MaxNoOfAttributes

    Defines the number of attributes that can be defined in the cluster.

    The default value is 1000, with the minimum possible value being 32. The maximum is 4294967039. Each attribute consumes around 200 bytes of storage per node due to the fact that all metadata is fully replicated on the servers.

    When setting MaxNoOfAttributes, it is important to prepare in advance for any ALTER TABLE statements that you might want to perform in the future. This is due to the fact, during the execution of ALTER TABLE on a Cluster table, 3 times the number of attributes as in the original table are used. For example, if a table requires 100 attributes, and you want to be able to alter it later, you need to set the value of MaxNoOfAttributes to 300. Assuming that you can create all desired tables without any problems, a good rule of thumb is to add two times the number of attributes in the largest table to MaxNoOfAttributes to be sure. You should also verify that this number is sufficient by trying an actual ALTER TABLE after configuring the parameter. If this is not successful, increase MaxNoOfAttributes by another multiple of the original value and test it again.

  • MaxNoOfTables

    A table object is allocated for each table, unique hash index, and ordered index. This parameter sets the maximum number of table objects for the cluster as a whole.

    For each attribute that has a BLOB data type an extra table is used to store most of the BLOB data. These tables also must be taken into account when defining the total number of tables.

    The default value of this parameter is 128. The minimum is 8 and the maximum is 1600. Each table object consumes approximately 20KB per node.

  • MaxNoOfOrderedIndexes

    For each ordered index in the cluster, an object is allocated describing what is being indexed and its storage segments. By default, each index so defined also defines an ordered index. Each unique index and primary key has both an ordered index and a hash index.

    The default value of this parameter is 128. Each object consumes approximately 10KB of data per node.

  • MaxNoOfUniqueHashIndexes

    For each unique index that is not a primary key, a special table is allocated that maps the unique key to the primary key of the indexed table. By default, an ordered index is also defined for each unique index. To prevent this, you must specify the USING HASH option when defining the unique index.

    The default value is 64. Each index consumes approximately 15KB per node.

  • MaxNoOfTriggers

    Internal update, insert, and delete triggers are allocated for each unique hash index. (This means that three triggers are created for each unique hash index.) However, an ordered index requires only a single trigger object. Backups also use three trigger objects for each normal table in the cluster.

    Replication between clusters also makes use of internal triggers.

    This parameter sets the maximum number of trigger objects in the cluster.

    The default value is 768.

  • MaxNoOfIndexes

    This parameter is deprecated in MySQL 5.1; you should use MaxNoOfOrderedIndexes and MaxNoOfUniqueHashIndexes instead.

    This parameter is used only by unique hash indexes. There needs to be one record in this pool for each unique hash index defined in the cluster.

    The default value of this parameter is 128.

Boolean Parameters

The behavior of data nodes is also affected by a set of [NDBD] parameters taking on boolean values. These parameters can each be specified as TRUE by setting them equal to 1 or Y, and as FALSE by setting them equal to 0 or N.

  • LockPagesInMainMemory

    For a number of operating systems, including Solaris and Linux, it is possible to lock a process into memory and so avoid any swapping to disk. This can be used to help guarantee the cluster's real-time characteristics.

    Beginning with MySQL 5.1.15, this parameter takes one of the integer values 0, 1, or 2, which act as follows:

    • 0: Disables locking. This is the default value.

    • 1: Performs the lock after allocating memory for the process.

    • 2: Performs the lock before memory for the process is allocated.

    Previously, this parameter was a Boolean. 0 or false was the default setting, and disabled locking. 1 or true enabled locking of the process after its memory was allocated. Important: Beginning with MySQL 5.1.15, it is no longer possible to use true or false for the value of this parameter; when upgrading from a previous version, you must change the value to 0, 1, or 2.

  • StopOnError

    This parameter specifies whether an ndbd process should exit or perform an automatic restart when an error condition is encountered.

    This feature is enabled by default.

  • Diskless

    It is possible to specify MySQL Cluster tables as diskless, meaning that tables are not checkpointed to disk and that no logging occurs. Such tables exist only in main memory. A consequence of using diskless tables is that neither the tables nor the records in those tables survive a crash. However, when operating in diskless mode, it is possible to run ndbd on a diskless computer.

    Important: This feature causes the entire cluster to operate in diskless mode.

    When this feature is enabled, Cluster online backup is disabled. In addition, a partial start of the cluster is not possible.

    Diskless is disabled by default.

  • RestartOnErrorInsert

    This feature is accessible only when building the debug version where it is possible to insert errors in the execution of individual blocks of code as part of testing.

    This feature is disabled by default.

Controlling Timeouts, Intervals, and Disk Paging

There are a number of [NDBD] parameters specifying timeouts and intervals between various actions in Cluster data nodes. Most of the timeout values are specified in milliseconds. Any exceptions to this are mentioned where applicable.

  • TimeBetweenWatchDogCheck

    To prevent the main thread from getting stuck in an endless loop at some point, a “watchdog” thread checks the main thread. This parameter specifies the number of milliseconds between checks. If the process remains in the same state after three checks, the watchdog thread terminates it.

    This parameter can easily be changed for purposes of experimentation or to adapt to local conditions. It can be specified on a per-node basis although there seems to be little reason for doing so.

    The default timeout is 4000 milliseconds (4 seconds).

  • StartPartialTimeout

    This parameter specifies how long the Cluster waits for all data nodes to come up before the cluster initialization routine is invoked. This timeout is used to avoid a partial Cluster startup whenever possible.

    The default value is 30000 milliseconds (30 seconds). 0 disables the timeout, in which case the cluster may start only if all nodes are available.

  • StartPartitionedTimeout

    If the cluster is ready to start after waiting for StartPartialTimeout milliseconds but is still possibly in a partitioned state, the cluster waits until this timeout has also passed.

    The default timeout is 60000 milliseconds (60 seconds).

  • StartFailureTimeout

    If a data node has not completed its startup sequence within the time specified by this parameter, the node startup fails. Setting this parameter to 0 (the default value) means that no data node timeout is applied.

    For nonzero values, this parameter is measured in milliseconds. For data nodes containing extremely large amounts of data, this parameter should be increased. For example, in the case of a data node containing several gigabytes of data, a period as long as 10–15 minutes (that is, 600000 to 1000000 milliseconds) might be required to to perform a node restart.

  • HeartbeatIntervalDbDb

    One of the primary methods of discovering failed nodes is by the use of heartbeats. This parameter states how often heartbeat signals are sent and how often to expect to receive them. After missing three heartbeat intervals in a row, the node is declared dead. Thus, the maximum time for discovering a failure through the heartbeat mechanism is four times the heartbeat interval.

    The default heartbeat interval is 1500 milliseconds (1.5 seconds). This parameter must not be changed drastically and should not vary widely between nodes. If one node uses 5000 milliseconds and the node watching it uses 1000 milliseconds, obviously the node will be declared dead very quickly. This parameter can be changed during an online software upgrade, but only in small increments.

  • HeartbeatIntervalDbApi

    Each data node sends heartbeat signals to each MySQL server (SQL node) to ensure that it remains in contact. If a MySQL server fails to send a heartbeat in time it is declared “dead,” in which case all ongoing transactions are completed and all resources released. The SQL node cannot reconnect until all activities initiated by the previous MySQL instance have been completed. The three-heartbeat criteria for this determination are the same as described for HeartbeatIntervalDbDb.

    The default interval is 1500 milliseconds (1.5 seconds). This interval can vary between individual data nodes because each data node watches the MySQL servers connected to it, independently of all other data nodes.

  • TimeBetweenLocalCheckpoints

    This parameter is an exception in that it does not specify a time to wait before starting a new local checkpoint; rather, it is used to ensure that local checkpoints are not performed in a cluster where relatively few updates are taking place. In most clusters with high update rates, it is likely that a new local checkpoint is started immediately after the previous one has been completed.

    The size of all write operations executed since the start of the previous local checkpoints is added. This parameter is also exceptional in that it is specified as the base-2 logarithm of the number of 4-byte words, so that the default value 20 means 4MB (4 × 220) of write operations, 21 would mean 8MB, and so on up to a maximum value of 31, which equates to 8GB of write operations.

    All the write operations in the cluster are added together. Setting TimeBetweenLocalCheckpoints to 6 or less means that local checkpoints will be executed continuously without pause, independent of the cluster's workload.

  • TimeBetweenGlobalCheckpoints

    When a transaction is committed, it is committed in main memory in all nodes on which the data is mirrored. However, transaction log records are not flushed to disk as part of the commit. The reasoning behind this behavior is that having the transaction safely committed on at least two autonomous host machines should meet reasonable standards for durability.

    It is also important to ensure that even the worst of cases — a complete crash of the cluster — is handled properly. To guarantee that this happens, all transactions taking place within a given interval are put into a global checkpoint, which can be thought of as a set of committed transactions that has been flushed to disk. In other words, as part of the commit process, a transaction is placed in a global checkpoint group. Later, this group's log records are flushed to disk, and then the entire group of transactions is safely committed to disk on all computers in the cluster.

    This parameter defines the interval between global checkpoints. The default is 2000 milliseconds.

  • TimeBetweenInactiveTransactionAbortCheck

    Timeout handling is performed by checking a timer on each transaction once for every interval specified by this parameter. Thus, if this parameter is set to 1000 milliseconds, every transaction will be checked for timing out once per second.

    The default value is 1000 milliseconds (1 second).

  • TransactionInactiveTimeout

    This parameter states the maximum time that is permitted to lapse between operations in the same transaction before the transaction is aborted.

    The default for this parameter is zero (no timeout). For a real-time database that needs to ensure that no transaction keeps locks for too long, this parameter should be set to a much smaller value. The unit is milliseconds.

  • TransactionDeadlockDetectionTimeout

    When a node executes a query involving a transaction, the node waits for the other nodes in the cluster to respond before continuing. A failure to respond can occur for any of the following reasons:

    • The node is “dead

    • The operation has entered a lock queue

    • The node requested to perform the action could be heavily overloaded.

    This timeout parameter states how long the transaction coordinator waits for query execution by another node before aborting the transaction, and is important for both node failure handling and deadlock detection. Setting it too high can cause a undesirable behavior in situations involving deadlocks and node failure.

    The default timeout value is 1200 milliseconds (1.2 seconds).

  • DiskSyncSize

    This is the maximum number of bytes to store before flushing data to a local checkpoint file.

    The default value is 4M (4 megabytes).

    This parameter was added in MySQL 5.1.12.

  • DiskCheckpointSpeed

    The amount of data,in bytes per second, that is sent to disk during a local checkpoint.

    The default value is 10M (10 megabytes per second).

    This parameter was added in MySQL 5.1.12.

  • DiskCheckpointSpeedInRestart

    The amount of data,in bytes per second, that is sent to disk during a local checkpoint as part of a restart operation.

    The default value is 100M (100 megabytes per second).

    This parameter was added in MySQL 5.1.12.

  • NoOfDiskPagesToDiskAfterRestartTUP

    When executing a local checkpoint, the algorithm flushes all data pages to disk. Merely doing so as quickly as possible without any moderation is likely to impose excessive loads on processors, networks, and disks. To control the write speed, this parameter specifies how many pages per 100 milliseconds are to be written. In this context, a “page” is defined as 8KB. This parameter is specified in units of 80KB per second, so , setting NoOfDiskPagesToDiskAfterRestartTUP to a value of 20 entails writing 1.6MB in data pages to disk each second during a local checkpoint. This value includes the writing of UNDO log records for data pages. That is, this parameter handles the limitation of writes from data memory. UNDO log records for index pages are handled by the parameter NoOfDiskPagesToDiskAfterRestartACC. (See the entry for IndexMemory for information about index pages.)

    In short, this parameter specifies how quickly to execute local checkpoints. It operates in conjunction with NoOfFragmentLogFiles, DataMemory, and IndexMemory.

    For more information about the interaction between these parameters and possible strategies for choosing appropriate values for them, see Section 15.4.6, “Configuring Parameters for Local Checkpoints”.

    The default value is 40 (3.2MB of data pages per second).

    Note: This parameter is deprecated as of MySQL 5.1.6. For MySQL 5.1.12 and later versions, use DiskCheckpointSpeed and DiskSyncSize.

  • NoOfDiskPagesToDiskAfterRestartACC

    This parameter uses the same units as NoOfDiskPagesToDiskAfterRestartTUP and acts in a similar fashion, but limits the speed of writing index pages from index memory.

    The default value of this parameter is 20 (1.6MB of index memory pages per second).

    Note: This parameter is deprecated as of MySQL 5.1.6. For MySQL 5.1.12 and later versions, use DiskCheckpointSpeed and DiskSyncSize.

  • NoOfDiskPagesToDiskDuringRestartTUP

    This parameter is used in a fashion similar to NoOfDiskPagesToDiskAfterRestartTUP and NoOfDiskPagesToDiskAfterRestartACC, only it does so with regard to local checkpoints executed in the node when a node is restarting. A local checkpoint is always performed as part of all node restarts. During a node restart it is possible to write to disk at a higher speed than at other times, because fewer activities are being performed in the node.

    This parameter covers pages written from data memory.

    The default value is 40 (3.2MB per second).

    Note: This parameter is deprecated as of MySQL 5.1.6. For MySQL 5.1.12 and later versions, use DiskCheckpointSpeedInRestart and DiskSyncSize.

  • NoOfDiskPagesToDiskDuringRestartACC

    Controls the number of index memory pages that can be written to disk during the local checkpoint phase of a node restart.

    As with NoOfDiskPagesToDiskAfterRestartTUP and NoOfDiskPagesToDiskAfterRestartACC, values for this parameter are expressed in terms of 8KB pages written per 100 milliseconds (80KB/second).

    The default value is 20 (1.6MB per second).

    Note: This parameter is deprecated as of MySQL 5.1.6. For MySQL 5.1.12 and later versions, use DiskCheckpointSpeedInRestart and DiskSyncSize.

  • ArbitrationTimeout

    This parameter specifies how long data nodes wait for a response from the arbitrator to an arbitration message. If this is exceeded, the network is assumed to have split.

    The default value is 1000 milliseconds (1 second).

Buffering and Logging


All update activities need to be logged. The REDO log makes it possible to replay these updates whenever the system is restarted. The NDB recovery algorithm uses a “fuzzy” checkpoint of the data together with the UNDO log, and then applies the REDO log to play back all changes up to the restoration point.

RedoBuffer sets the size of the buffer inwhich the REDO log is written, and is 8MB by default. This buffer is used as a front end to the file system when writing REDO log records to disk. If this buffer is too small, the NDB storage engine issues error code 1221 (REDO log buffers overloaded).

The minimum value is 1MB. If the node is running in diskless mode, these parameters can be set to their minimum values without penalty due to the fact that disk writes are “faked” by the NDB storage engine's filesystem abstraction layer.

Important: It is not safe to decrease the value of this parameter during a rolling restart.

Note: The UndoIndexBuffer and UndoDataBuffer parameters which appeared in earlier versions of MySQL Cluster are no longer necessary (or even present) in MySQL 5.1.

Controlling Log Messages

In managing the cluster, it is very important to be able to control the number of log messages sent for various event types to stdout. For each event category, there are 16 possible event levels (numbered 0 through 15). Setting event reporting for a given event category to level 15 means all event reports in that category are sent to stdout; setting it to 0 means that there will be no event reports made in that category.

By default, only the startup message is sent to stdout, with the remaining event reporting level defaults being set to 0. The reason for this is that these messages are also sent to the management server's cluster log.

An analogous set of levels can be set for the management client to determine which event levels to record in the cluster log.

  • LogLevelStartup

    The reporting level for events generated during startup of the process.

    The default level is 1.

  • LogLevelShutdown

    The reporting level for events generated as part of graceful shutdown of a node.

    The default level is 0.

  • LogLevelStatistic

    The reporting level for statistical events such as number of primary key reads, number of updates, number of inserts, information relating to buffer usage, and so on.

    The default level is 0.

  • LogLevelCheckpoint

    The reporting level for events generated by local and global checkpoints.

    The default level is 0.

  • LogLevelNodeRestart

    The reporting level for events generated during node restart.

    The default level is 0.

  • LogLevelConnection

    The reporting level for events generated by connections between cluster nodes.

    The default level is 0.

  • LogLevelError

    The reporting level for events generated by errors and warnings by the cluster as a whole. These errors do not cause any node failure but are still considered worth reporting.

    The default level is 0.

  • LogLevelInfo

    The reporting level for events generated for information about the general state of the cluster.

    The default level is 0.

Backup Parameters

The [NDBD] parameters discussed in this section define memory buffers set aside for execution of online backups.

  • BackupDataBufferSize

    In creating a backup, there are two buffers used for sending data to the disk. The backup data buffer is used to fill in data recorded by scanning a node's tables. Once this buffer has been filled to the level specified as BackupWriteSize (see below), the pages are sent to disk. While flushing data to disk, the backup process can continue filling this buffer until it runs out of space. When this happens, the backup process pauses the scan and waits until some disk writes have completed freed up memory so that scanning may continue.

    The default value is 2MB.

  • BackupLogBufferSize

    The backup log buffer fulfills a role similar to that played by the backup data buffer, except that it is used for generating a log of all table writes made during execution of the backup. The same principles apply for writing these pages as with the backup data buffer, except that when there is no more space in the backup log buffer, the backup fails. For that reason, the size of the backup log buffer must be large enough to handle the load caused by write activities while the backup is being made. See Section 15.8.4, “Configuration for Cluster Backup”.

    The default value for this parameter should be sufficient for most applications. In fact, it is more likely for a backup failure to be caused by insufficient disk write speed than it is for the backup log buffer to become full. If the disk subsystem is not configured for the write load caused by applications, the cluster is unlikely to be able to perform the desired operations.

    It is preferable to configure cluster nodes in such a manner that the processor becomes the bottleneck rather than the disks or the network connections.

    The default value is 2MB.

  • BackupMemory

    This parameter is simply the sum of BackupDataBufferSize and BackupLogBufferSize.

    The default value is 2MB + 2MB = 4MB.

    Important: If BackupDataBufferSize and BackupLogBufferSize taken together exceed 4MB, then this parameter must be set explicitly in the config.ini file to their sum.

  • BackupWriteSize

    This parameter specifies the default size of messages written to disk by the backup log and backup data buffers.

    The default value is 32KB.

  • BackupMaxWriteSize

    This parameter specifies the maximum size of messages written to disk by the backup log and backup data buffers.

    The default value is 256KB.

Important: When specifying these parameters, the following relationships must hold true. Otherwise, the data node will be unable to start.

  • BackupDataBufferSize >= BackupWriteSize + 188KB

  • BackupLogBufferSize >= BackupWriteSize + 16KB

  • BackupMaxWriteSize >= BackupWriteSize Defining SQL and Other API Nodes

The [MYSQLD] and [API] sections in the config.ini file define the behavior of the MySQL servers (SQL nodes) and other applications (API nodes) used to access cluster data. None of the parameters shown is required. If no computer or host name is provided, any host can use this SQL or API node.

Generally speaking, a [MYSQLD] section is used to indicate a MySQL server providing an SQL interface to the cluster, and an [API] section is used for applications other than mysqld processes accessing cluster data, but the two designations are actually synonomous; you can, for instance, list parameters for a MySQL server acting as an SQL node in an [API] section.

  • Id

    The Id value is used to identify the node in all cluster internal messages. It must be an integer in the range 1 to 63 inclusive, and must be unique among all node IDs within the cluster.

  • ExecuteOnComputer

    This refers to the Id set for one of the computers (hosts) defined in a [COMPUTER] section of the configuration file.

  • HostName

    Specifying this parameter defines the hostname of the computer on which the SQL node (API node) is to reside. To specify a hostname other than localhost, either this parameter or ExecuteOnComputer is required.

  • ArbitrationRank

    This parameter defines which nodes can act as arbitrators. Both MGM nodes and SQL nodes can be arbitrators. A value of 0 means that the given node is never used as an arbitrator, a value of 1 gives the node high priority as an arbitrator, and a value of 2 gives it low priority. A normal configuration uses the management server as arbitrator, setting its ArbitrationRank to 1 (the default) and those for all SQL nodes to 0.

  • ArbitrationDelay

    Setting this parameter to any other value than 0 (the default) means that responses by the arbitrator to arbitration requests will be delayed by the stated number of milliseconds. It is usually not necessary to change this value.

  • BatchByteSize

    For queries that are translated into full table scans or range scans on indexes, it is important for best performance to fetch records in properly sized batches. It is possible to set the proper size both in terms of number of records (BatchSize) and in terms of bytes (BatchByteSize). The actual batch size is limited by both parameters.

    The speed at which queries are performed can vary by more than 40% depending upon how this parameter is set. In future releases, MySQL Server will make educated guesses on how to set parameters relating to batch size, based on the query type.

    This parameter is measured in bytes and by default is equal to 32KB.

  • BatchSize

    This parameter is measured in number of records and is by default set to 64. The maximum size is 992.

  • MaxScanBatchSize

    The batch size is the size of each batch sent from each data node. Most scans are performed in parallel to protect the MySQL Server from receiving too much data from many nodes in parallel; this parameter sets a limit to the total batch size over all nodes.

    The default value of this parameter is set to 256KB. Its maximum size is 16MB.

You can obtain some information from a MySQL server running as a Cluster SQL node using SHOW STATUS in the mysql client, as shown here:

mysql> SHOW STATUS LIKE 'ndb%';
| Variable_name               | Value         |
| Ndb_cluster_node_id         | 5             | 
| Ndb_config_from_host        | | 
| Ndb_config_from_port        | 1186          | 
| Ndb_number_of_storage_nodes | 4             | 
4 rows in set (0.02 sec)

For information about these Cluster system status variables, see Section 5.2.5, “Status Variables”. Cluster TCP/IP Connections

TCP/IP is the default transport mechanism for establishing connections in MySQL Cluster. It is normally not necessary to define connections because Cluster automatically set ups a connection between each of the data nodes, between each data node and all MySQL server nodes, and between each data node and the management server. (For one exception to this rule, see Section, “TCP/IP Connections Using Direct Connections”.) [TCP] sections in the config.ini file explicitly define TCP/IP connections between nodes in the cluster.

It is only necessary to define a connection to override the default connection parameters. In that case, it is necessary to define at least NodeId1, NodeId2, and the parameters to change.

It is also possible to change the default values for these parameters by setting them in the [TCP DEFAULT] section.

  • NodeId1, NodeId2

    To identify a connection between two nodes it is necessary to provide their node IDs in the [TCP] section of the configuration file. These are the same unique Id values for each of these nodes as described in Section, “Defining SQL and Other API Nodes”.

  • SendBufferMemory

    TCP transporters use a buffer to store all messages before performing the send call to the operating system. When this buffer reaches 64KB its contents are sent; these are also sent when a round of messages have been executed. To handle temporary overload situations it is also possible to define a bigger send buffer. The default size of the send buffer is 256KB.

  • SendSignalId

    To be able to retrace a distributed message datagram, it is necessary to identify each message. When this parameter is set to Y, message IDs are transported over the network. This feature is disabled by default in production builds, and enabled in -debug builds.

  • Checksum

    This parameter is a boolean parameter (enabled by setting it to Y or 1, disabled by setting it to N or 0). It is disabled by default. When it is enabled, checksums for all messages are calculated before they placed in the send buffer. This feature ensures that messages are not corrupted while waiting in the send buffer, or by the transport mechanism.

  • PortNumber (OBSOLETE)

    This formerly specified the port number to be used for listening for connections from other nodes. This parameter should no longer be used.

  • ReceiveBufferMemory

    Specifies the size of the buffer used when receiving data from the TCP/IP socket. There is seldom any need to change this parameter from its default value of 64KB, except possibly to save memory. TCP/IP Connections Using Direct Connections

Setting up a cluster using direct connections between data nodes requires specifying explicitly the crossover IP addresses of the data nodes so connected in the [TCP] section of the cluster config.ini file.

In the following example, we envision a cluster with at least four hosts, one each for a management server, an SQL node, and two data nodes. The cluster as a whole resides on the 172.23.72.* subnet of a LAN. In addition to the usual network connections, the two data nodes are connected directly using a standard crossover cable, and communicate with one another directly using IP addresses in the 1.1.0.* address range as shown:

# Management Server

# SQL Node

# Data Nodes


# TCP/IP Connections

The HostNameN parameter, where N is an integer, is used only when specifying direct TCP/IP connections.

The use of direct connections between data nodes can improve the cluster's overall efficiency by allowing the data nodes to bypass an Ethernet device such as a switch, hub, or router, thus cutting down on the cluster's latency. It is important to note that to take the best advantage of direct connections in this fashion with more than two data nodes, you must have a direct connection between each data node and every other data node in the same node group. Shared-Memory Connections

MySQL Cluster attempts to use the shared memory transporter and configure it automatically where possible. [SHM] sections in the config.ini file explicitly define shared-memory connections between nodes in the cluster. When explicitly defining shared memory as the connection method, it is necessary to define at least NodeId1, NodeId2 and ShmKey. All other parameters have default values that should work well in most cases.

Important: SHM functionality is considered experimental only. It is not officially supported in any MySQL release series up to and including 5.1. This means that you must determine for yourself or by using our free resources (forums, mailing lists) whether it can be made to work correctly in your specific case.

  • NodeId1, NodeId2

    To identify a connection between two nodes it is necessary to provide node identifiers for each of them, as NodeId1 and NodeId2.

  • ShmKey

    When setting up shared memory segments, a node ID, expressed as an integer, is used to identify uniquely the shared memory segment to use for the communication. There is no default value.

  • ShmSize

    Each SHM connection has a shared memory segment where messages between nodes are placed by the sender and read by the reader. The size of this segment is defined by ShmSize. The default value is 1MB.

  • SendSignalId

    To retrace the path of a distributed message, it is necessary to provide each message with a unique identifier. Setting this parameter to Y causes these message IDs to be transported over the network as well. This feature is disabled by default in production builds, and enabled in -debug builds.

  • Checksum

    This parameter is a boolean (Y/N) parameter which is disabled by default. When it is enabled, checksums for all messages are calculated before being placed in the send buffer.

    This feature prevents messages from being corrupted while waiting in the send buffer. It also serves as a check against data being corrupted during transport. SCI Transport Connections

[SCI] sections in the config.ini file explicitly define SCI (Scalable Coherent Interface) connections between cluster nodes. Using SCI transporters in MySQL Cluster is supported only when the MySQL binaries are built using --with-ndb-sci=/your/path/to/SCI. The path should point to a directory that contains at a minimum lib and include directories containing SISCI libraries and header files. (See Section 15.12, “Using High-Speed Interconnects with MySQL Cluster” for more information about SCI.)

In addition, SCI requires specialized hardware.

It is strongly recommended to use SCI Transporters only for communication between ndbd processes. Note also that using SCI Transporters means that the ndbd processes never sleep. For this reason, SCI Transporters should be used only on machines having at least two CPUs dedicated for use by ndbd processes. There should be at least one CPU per ndbd process, with at least one CPU left in reserve to handle operating system activities.

  • NodeId1, NodeId2

    To identify a connection between two nodes it is necessary to provide node identifiers for each of them, as NodeId1 and NodeId2.

  • Host1SciId0

    This identifies the SCI node ID on the first Cluster node (identified by NodeId1).

  • Host1SciId1

    It is possible to set up SCI Transporters for failover between two SCI cards which then should use separate networks between the nodes. This identifies the node ID and the second SCI card to be used on the first node.

  • Host2SciId0

    This identifies the SCI node ID on the second Cluster node (identified by NodeId2).

  • Host2SciId1

    When using two SCI cards to provide failover, this parameter identifies the second SCI card to be used on the second node.

  • SharedBufferSize

    Each SCI transporter has a shared memory segment used for communication between the two nodes. Setting the size of this segment to the default value of 1MB should be sufficient for most applications. Using a smaller value can lead to problems when performing many parallel inserts; if the shared buffer is too small, this can also result in a crash of the ndbd process.

  • SendLimit

    A small buffer in front of the SCI media stores messages before transmitting them over the SCI network. By default, this is set to 8KB. Our benchmarks show that performance is best at 64KB but 16KB reaches within a few percent of this, and there was little if any advantage to increasing it beyond 8KB.

  • SendSignalId

    To trace a distributed message it is necessary to identify each message uniquely. When this parameter is set to Y, message IDs are transported over the network. This feature is disabled by default in production builds, and enabled in -debug builds.

  • Checksum

    This parameter is a boolean value, and is disabled by default. When Checksum is enabled, checksums are calculated for all messages before they are placed in the send buffer. This feature prevents messages from being corrupted while waiting in the send buffer. It also serves as a check against data being corrupted during transport.

15.4.5. Overview of Cluster Configuration Parameters

The next three sections provide summary tables of MySQL Cluster configuration parameters used in the config.ini file to govern the cluster's functioning. Each table lists the parameters for one of the Cluster node process types (ndbd, ndb_mgmd, and mysqld), and includes the parameter's type as well as its default, mimimum, and maximum values as applicable.

It is also stated what type of restart is required (node restart or system restart) — and whether the restart must be done with --initial — to change the value of a given configuration parameter. This information is provided in each table's Restart Type column, which contains one of the values shown in this list:

  • N: Node Restart

  • IN: Initial Node Restart

  • S: System Restart

  • IS: Initial System Restart

When performing a node restart or an initial node restart, all of the cluster's data nodes must be restarted in turn (also referred to as a rolling restart). It is possible to update cluster configuration parameters marked N or IN online — that is, without shutting down the cluster — in this fashion. An initial node restart requires restarting each ndbd process with the --initial option.

A system restart requires a complete shutdown and restart of the entire cluster. An initial system restart requires taking a backup of the cluster, wiping the cluster filesystem after shutdown, and then restoring from the backup following the restart.

In any cluster restart, all of the cluster's management servers must be restarted in order for them to read the updated configuration parameter values.

Important: Values for numeric cluster parameters can generally be increased without any problems, although it is advisable to do so progressively, making such adjustments in relatively small increments. However, decreasing the values of such parameters — particularly those relating to memory usage and disk space — is not to be undertaken lightly, and it is recommended that you do so only following careful planning and testing. In addition, it is the generally the case that parameters relating to memory and disk usage which can be raised using a simple node restart require an initial node restart to be lowered.

Because some of these parameters can be used for configuring more than one type of cluster node, they may appear in more than one of the tables.

(Note that 4294967039 — which often appears as a maximum value in these tables — is equal to 232 – 28 – 1.) Data Node Configuration Parameters

The following table provides information about parameters used in the [NDBD] or [NDB_DEFAULT] sections of a config.ini file for configuring MySQL Cluster data nodes. For detailed descriptions and other additional information about each of these parameters, see Section, “Defining Data Nodes”.

Restart Type Column Values

  • N: Node Restart

  • IN: Initial Node Restart

  • S: System Restart

  • IS: Initial System Restart

See Section 15.4.5, “Overview of Cluster Configuration Parameters”, for additional explanations of these abbreviations.

Parameter NameType/UnitsDefault ValueMinimum ValueMaximum ValueRestart Type
DataMemorybytes80M1M1024G (subject to available system RAM and size of IndexMemory)N
Disklesstrue|false (1|0)001IS
FileSystemPathstringvalue specified for DataDirN/AN/AIN
IndexMemorybytes18M1M1024G (subject to available system RAM and size of DataMemory)N
LockPagesInMainMemoryAs of MySQL 5.1.15: integer; previously: true|false (1|0)001N
MaxNoOfIndexes (DEPRECATED — use MaxNoOfOrderedIndexes or MaxNoOfUniqueHashIndexes instead)integer12804294967039N
NoOfDiskPagesToDiskAfterRestartACC (DEPRECATED as of MySQL 5.1.6)integer (number of 8KB pages per 100 milliseconds)20 (= 20 * 80KB = 1.6MB/second)14294967039N
NoOfDiskPagesToDiskAfterRestartTUP (DEPRECATED as of MySQL 5.1.6)integer (number of 8KB pages per 100 milliseconds)40 (= 40 * 80KB = 3.2MB/second)14294967039N
NoOfDiskPagesToDiskDuringRestartACC (DEPRECATED as of MySQL 5.1.6)integer (number of 8KB pages per 100 milliseconds)20 (= 20 * 80KB = 1.6MB/second)14294967039N
NoOfDiskPagesToDiskDuringRestartTUP (DEPRECATED as of MySQL 5.1.6)integer (number of 8KB pages per 100 milliseconds)40 (= 40 * 80KB = 3.2MB/second)14294967039N
DiskCheckpointSpeed (added in MySQL 5.1.12)integer (number of bytes per second)10M1M4294967039N
DiskCheckpointSpeedInRestart (added in MySQL 5.1.12)integer (number of bytes per second)100M1M4294967039N
DiskSyncSize (added in MySQL 5.1.12)integer (number of bytes)4M32K4294967039N
RestartOnErrorInsert (DEBUG BUILDS ONLY)true|false (1|0)001N
ServerPort (OBSOLETE)integer118604294967039N
StopOnErrortrue|false (1|0)101N
TimeBetweenLocalCheckpointsinteger (number of 4-byte words as a base-2 logarithm)20 (= 4 * 220 = 4MB write operations)031N
TransactionInactiveTimeoutmilliseconds004294967039N Management Node Configuration Parameters

The following table provides information about parameters used in the [NDB_MGMD] or [MGM] sections of a config.ini file for configuring MySQL Cluster management nodes. For detailed descriptions and other additional information about each of these parameters, see Section, “Defining the Management Server”.

Restart Type Column Values

  • N: Node Restart

  • IN: Initial Node Restart

  • S: System Restart

  • IS: Initial System Restart

See Section 15.4.5, “Overview of Cluster Configuration Parameters”, for additional explanations of these abbreviations.

Parameter NameType/UnitsDefault ValueMinimum ValueMaximum ValueRestart Type
DataDirstring./ (ndb_mgmd directory)N/AN/AIN
LogDestinationCONSOLE, SYSLOG, or FILEFILE (see Section, “Defining the Management Server”)N/AN/AN SQL Node and API Node Configuration Parameters

The following table provides information about parameters used in the [SQL] and [API] sections of a config.ini file for configuring MySQL Cluster SQL nodes and API nodes. For detailed descriptions and other additional information about each of these parameters, see Section, “Defining SQL and Other API Nodes”.

Restart Type Column Values

  • N: Node Restart

  • IN: Initial Node Restart

  • S: System Restart

  • IS: Initial System Restart

See Section 15.4.5, “Overview of Cluster Configuration Parameters”, for additional explanations of these abbreviations.

Parameter NameType/UnitsDefault ValueMinimum ValueMaximum ValueRestart Type

15.4.6. Configuring Parameters for Local Checkpoints

The parameters discussed in Logging and Checkpointing and in Data Memory, Index Memory, and String Memory that are used to configure local checkpoints for a MySQL Cluster do not exist in isolation, but rather are very much interdepedent on each other. In this section, we illustrate how these parameters — including DataMemory, IndexMemory, NoOfDiskPagesToDiskAfterRestartTUP, NoOfDiskPagesToDiskAfterRestartACC, and NoOfFragmentLogFiles — relate to one another in a working Cluster.

Important: The parameters NoOfDiskPagesToDiskAfterRestartTUP and NoOfDiskPagesToDiskAfterRestartACC were deprecated in MySQL 5.1.6. From MySQL 5.1.6 through 5.1.11, disk writes during LCPs took place at the maximum speed possible. Beginning with MySQL 5.1.12, the speed and throughput for LCPs are controlled using the parameters DiskSyncSize, DiskCheckpointSpeed, and DiskCheckpointSpeedInRestart. See Section, “Defining Data Nodes”.

In this example, we assume that our application performs the following numbers of types of operations per hour:

  • 50000 selects

  • 15000 inserts

  • 15000 updates

  • 15000 deletes

We also make the following assumptions about the data used in the application:

  • We are working with a single table having 40 columns.

  • Each column can hold up to 32 bytes of data.

  • A typical UPDATE run by the application affects the values of 5 columns.

  • No NULL values are inserted by the application.

A good starting point is to determine the amount of time that should elapse between local checkpoints (LCPs). It worth noting that, in the event of a system restart, it takes 40-60 percent of this interval to execute the REDO log — for example, if the time between LCPs is 5 minutes (300 seconds), then it should take 2 to 3 minutes (120 to 180 seconds) for the REDO log to be read.

The maximum amount of data per node can be assumed to be the size of the DataMemory parameter. In this example, we assume that this is 2 GB. The NoOfDiskPagesToDiskAfterRestartTUP parameter represents the amount of data to be checkpointed per unit time — however, this parameter is actually expressed as the number of 8K memory pages to be checkpointed per 100 milliseconds. 2 GB per 300 seconds is approximately 6.8 MB per second, or 700 KB per 100 milliseconds, which works out to roughly 85 pages per 100 milliseconds.

Similarly, we can calculate NoOfDiskPagesToDiskAfterRestartACC in terms of the time for local checkpoints and the amount of memory required for indexes — that is, the IndexMemory. Assuming that we allow 512 MB for indexes, this works out to approximately 20 8-KB pages per 100 milliseconds for this parameter.

Next, we need to determine the number of REDO log files required — that is, fragment log files — the corresponding parameter being NoOfFragmentLogFiles. We need to make sure that there are sufficient REDO log files for keeping records for at least 3 local checkpoints. In a production setting, there are always uncertainties — for instance, we cannot be sure that disks always operate at top speed or with maximum throughput. For this reason, it is best to err on the side of caution, so we double our requirement and calculate a number of fragment log files which should be enough to keep records covering 6 local checkpoints.

It is also important to remember that the disk also handles writes to the REDO log and UNDO log, so if you find that the amount of data being written to disk as detemined by the values of NoOfDiskPagesToDiskAfterRestartACC and NoOfDiskPagesToDiskAfterRestartTUP is approaching the amount of disk bandwidth available, you may wish to increase the time between local checkpoints.

Given 5 minutes (300 seconds) per local checkpoint, this means that we need to support writing log records at maximum speed for 6 * 300 = 1800 seconds. The size of a REDO log record is 72 bytes plus 4 bytes per updated column value plus the maximum size of the updated column, and there is one REDO log record for each table record updated in a transaction, on each node where the data reside. Using the numbers of operations set out previously in this section, we derive the following:

  • 50000 select operations per hour yields 0 log records (and thus 0 bytes), since SELECT statements are not recorded in the REDO log.

  • 15000 DELETE statements per hour is approximately 5 delete operations per second. (Since we wish to be conservative in our estimate, we round up here and in the following calculations.) No columns are updated by deletes, so these statements consume only 5 operations * 72 bytes per operation = 360 bytes per second.

  • 15000 UPDATE statements per hour is roughly the same as 5 updates per second. Each update uses 72 bytes, plus 4 bytes per column * 5 columns updated, plus 32 bytes per column * 5 columns — this works out to 72 + 20 + 160 = 252 bytes per operation, and multiplying this by 5 operation per second yields 1260 bytes per second.

  • 15000 INSERT statements per hour is equivalent to 5 insert operations per second. Each insert requires REDO log space of 72 bytes, plus 4 bytes per record * 40 columns, plus 32 bytes per column * 40 columns, which is 72 + 160 + 1280 = 1512 bytes per operation. This times 5 operations per second yields 7560 bytes per second.

So the total number of REDO log bytes being written per second is approximately 0 + 360 + 1260 + 7560 = 9180 bytes. Mutiplied by 1800 seconds, this yields 16524000 bytes required for REDO logging, or approximately 15.75 MB. The unit used for NoOfFragmentLogFiles represents a set of 4 16-MB log files — that is, 64 MB. Thus, the minimum value (3) for this parameter is sufficient for the scenario envisioned in this example, since 3 times 64 = 192 MB, or about 12 times what is required; the default value of 8 (or 512 MB) is more than ample in this case.

A copy of each altered table record is kept in the UNDO log. In the scenario discussed above, the UNDO log would not require any more space than what is provided by the default seetings. However, given the size of disks, it is sensible to allocate at least 1 GB for it.

15.5. Upgrading and Downgrading MySQL Cluster

This portion of the MySQL Cluster chapter covers upgrading and downgrading a MySQL Cluster from one MySQL release to another. It discusses different types of Cluster upgrades and downgrades, and provides a Cluster upgrade/downgrade compatibility matrix (see Section 15.5.2, “Cluster Upgrade and Downgrade Compatibility”).

Important: You are expected already to be familiar with installing and configuring a MySQL Cluster prior to attempting an upgrade or downgrade. See Section 15.4, “MySQL Cluster Configuration”.

This section remains in development, and continues to be updated and expanded.

15.5.1. Performing a Rolling Restart of the Cluster

This section discusses how to perform a rolling restart of a MySQL Cluster installation, so called because it involves stopping and starting (or restarting) each node in turn, so that the cluster itself remains operational. This is often done as part of a rolling upgrade or rolling downgrade, where high availability of the cluster is mandatory and no downtime of the cluster as a whole is permissible. Where we refer to upgrades, the information provided here also generally applies to downgrades as well.

There are a number of reasons why a rolling restart might be desirable:

  • Cluster Configuration Change: To make a change in the cluster's configuration, such as adding an SQL node to the cluster, or setting a configuration parameter to a new value.

  • Cluster Software Upgrade/Downgrade: To upgrade the cluster to a newer version of the MySQL Cluster software (or to downgrade it to an older version). This is usually referred to as a “rolling upgrade” (or “rolling downgrade”, when reverting to an older version of MySQL Cluster).

  • Change on Node Host: To make changes in the hardware or operating system on which one or more cluster nodes are running

  • Cluster Reset: To reset the cluster because it has reached an undesirable state

  • Freeing of Resources: To allow memory allocated to a table by successive INSERT and DELETE operations to be freed for re-use by other Cluster tables

The process for performing a rolling restart may be generalised as follows:

  1. Stop all cluster management nodes (ndb_mgmd processes), reconfigure them, then restart them

  2. Stop, reconfigure, then restart each cluster data node (ndbd process) in turn

  3. Stop, reconfigure, then restart each cluster SQL node (mysqld process) in turn

The specifics for implementing a particular rolling upgrade depend upon the actual changes being made. A more detailed view of the process is presented here:

MySQL Cluster Rolling Restarts (By

In the previous diagram, Stop and Start steps indicate that the process must be stopped completely using a shell command (such as kill on most Unix systems) or the management client STOP command, then started again from a system shell by invoking the ndbd or ndb_mgmd executable as appropriate. Restart indicates the process may be restarted using the ndb_mgm management client RESTART command.

15.5.2. Cluster Upgrade and Downgrade Compatibility

This section provides information regarding Cluster software and table file compatibility between differing versions of the MySQL Server for purposes of performing upgrades and downgrades.

Important: Only compatibility between MySQL versions with regard to NDB Cluster is taken into account in this section, and there are likely other issues to be considered. As with any other MySQL software upgrade or downgrade, you are strongly encouraged to review the relevant portions of the MySQL Manual for the MySQL versions from which and to which you intend to migrate, before attempting an upgrade or downgrade of the MySQL Cluster software. See Section 2.11, “Upgrading MySQL”.

The following table shows Cluster upgrade and downgrade compatibility between different versions of the MySQL Server.

MySQL Cluster upgrade/downgrade
          compatibility, by MySQL server version


  • 4.1 Series:

    You cannot upgrade directly from 4.1.8 to 4.1.10 (or newer); you must first upgrade from 4.1.8 to 4.1.9, then upgrade to 4.1.10. Similarly, you cannot downgrade directly from 4.1.10 (or newer) to 4.1.8; you must first downgrade from 4.1.10 to 4.1.9, then downgrade from 4.1.9 to 4.1.8.

    If you wish to upgrade a MySQL Cluster to 4.1.15, you must upgrade to 4.1.14 first, and you must upgrade to 4.1.15 before upgrading to 4.1.16 or newer.

    Cluster downgrades from 4.1.15 to 4.1.14 (or earlier versions) are not supported.

    Cluster upgrades from MySQL Server versions previous to 4.1.8 are not supported; when upgrading from these, you must dump all NDB tables, install the new version of the software, and then reload the tables from the dump.

  • 5.0 Series:

    MySQL 5.0.2 was the first public release in this series.

    Cluster downgrades from MySQL 5.0 to MySQL 4.1 are not supported.

    Cluster downgrades from 5.0.12 to 5.0.11 (or earlier) are not supported.

    You cannot restore with ndb_restore to a MySQL 5.0 Cluster using a backup made from a Cluster running MySQL 5.1. You must use mysqldump in such cases.

    There was no public release for MySQL 5.0.23.

  • 5.1 Series:

    MySQL 5.1.3 was the first public release in this series.

    You cannot downgrade a MySQL 5.1.6 or later Cluster using Disk Data tables to MySQL 5.1.5 or earlier unless you convert all such tables to in-memory Cluster tables first.

    MySQL 5.1.8 and MySQL 5.1.10 were not released.

    Online cluster upgrades and downgrades between MySQL 5.1.11 (or an earlier version) and 5.1.12 (or a later version) are not possible due to major changes in the cluster filesystem. In such cases, you must perform a backup or dump, upgrade (or downgrade) the software, start each data node with --initial, and then restore from the backup or dump. You can use NDB backup/restore or mysqldump for this purpose.

15.6. Process Management in MySQL Cluster

Understanding how to manage MySQL Cluster requires a knowledge of four essential processes. In the next few sections of this chapter, we cover the roles played by these processes in a cluster, how to use them, and what startup options are available for each of them:

15.6.1. MySQL Server Process Usage for MySQL Cluster

mysqld is the traditional MySQL server process. To be used with MySQL Cluster, mysqld needs to be built with support for the NDB Cluster storage engine, as it is in the precompiled binaries available from If you build MySQL from source, you must invoke configure with the --with-ndbcluster option to enable NDB Cluster storage engine support.

If the mysqld binary has been built with Cluster support, the NDB Cluster storage engine is still disabled by default. You can use either of two possible options to enable this engine:

  • Use --ndbcluster as a startup option on the command line when starting mysqld.

  • Insert a line containing ndbcluster in the [mysqld] section of your my.cnf file.

An easy way to verify that your server is running with the NDB Cluster storage engine enabled is to issue the SHOW ENGINES statement in the MySQL Monitor (mysql). You should see the value YES as the Support value in the row for NDBCLUSTER. If you see NO in this row or if there is no such row displayed in the output, you are not running an NDB-enabled version of MySQL. If you see DISABLED in this row, you need to enable it in either one of the two ways just described.

To read cluster configuration data, the MySQL server requires at a minimum three pieces of information:

  • The MySQL server's own cluster node ID

  • The hostname or IP address for the management server (MGM node)

  • The number of the TCP/IP port on which it can connect to the management server

Node IDs can be allocated dynamically, so it is not strictly necessary to specify them explicitly.

The mysqld parameter ndb-connectstring is used to specify the connectstring either on the command line when starting mysqld or in my.cnf. The connectstring contains the hostname or IP address where the management server can be found, as well as the TCP/IP port it uses.

In the following example, is the host where the management server resides, and the management server listens for cluster messages on port 1186:

shell> mysqld

See Section, “The Cluster connectstring, for more information on connectstrings.

Given this information, the MySQL server will be a full participant in the cluster. (We sometimes refer to a mysqld process running in this manner as an SQL node.) It will be fully aware of all cluster data nodes as well as their status, and will establish connections to all data nodes. In this case, it is able to use any data node as a transaction coordinator and to read and update node data.

You can see in the mysql client whether a MySQL server is connected to the cluster using SHOW PROCESSLIST. If the MySQL server is connected to the cluster, and you have the PROCESS privilege, then the first row of the output is as shown here:

*************************** 1. row ***************************
     Id: 1
   User: system user
Command: Daemon
   Time: 1
  State: Waiting for event from ndbcluster
   Info: NULL

15.6.2. ndbd, the Storage Engine Node Process

ndbd is the process that is used to handle all the data in tables using the NDB Cluster storage engine. This is the process that empowers a data node to accomplish distributed transaction handling, node recovery, checkpointing to disk, online backup, and related tasks.

In a MySQL Cluster, a set of ndbd processes cooperate in handling data. These processes can execute on the same computer (host) or on different computers. The correspondences between data nodes and Cluster hosts is completely configurable.

ndbd generates a set of log files which are placed in the directory specified by DataDir in the config.ini configuration file. These log files are listed below. Note that node_id represents the node's unique identifier. For example, ndb_2_error.log is the error log generated by the data node whose node ID is 2.

  • ndb_node_id_error.log is a file containing records of all crashes which the referenced ndbd process has encountered. Each record in this file contains a brief error string and a reference to a trace file for this crash. A typical entry in this file might appear as shown here:

    Date/Time: Saturday 30 July 2004 - 00:20:01
    Type of error: error
    Message: Internal program error (failed ndbrequire)
    Fault ID: 2341
    Problem data: DbtupFixAlloc.cpp
    Object of reference: DBTUP (Line: 173)
    ProgramName: NDB Kernel
    ProcessID: 14909
    TraceFile: ndb_2_trace.log.2

    Note: It is very important to be aware that the last entry in the error log file is not necessarily the newest one (nor is it likely to be). Entries in the error log are not listed in chronological order; rather, they correspond to the order of the trace files as determined in the file (see below). Error log entries are thus overwritten in a cyclical and not sequential fashion.

  • ndb_node_id_trace.log.trace_id is a trace file describing exactly what happened just before the error occurred. This information is useful for analysis by the MySQL Cluster development team.

    It is possible to configure the number of these trace files that will be created before old files are overwritten. trace_id is a number which is incremented for each successive trace file.

  • is the file that keeps track of the next trace file number to be assigned.

  • ndb_node_id_out.log is a file containing any data output by the ndbd process. This file is created only if ndbd is started as a daemon, which is the default behavior.

  • is a file containing the process ID of the ndbd process when started as a daemon. It also functions as a lock file to avoid the starting of nodes with the same identifier.

  • ndb_node_id_signal.log is a file used only in debug versions of ndbd, where it is possible to trace all incoming, outgoing, and internal messages with their data in the ndbd process.

It is recommended not to use a directory mounted through NFS because in some environments this can cause problems whereby the lock on the .pid file remains in effect even after the process has terminated.

To start ndbd, it may also be necessary to specify the hostname of the management server and the port on which it is listening. Optionally, one may also specify the node ID that the process is to use.

shell> ndbd --connect-string="nodeid=2;"

See Section, “The Cluster connectstring, for additional information about this issue. Section 15.6.5, “Command Options for MySQL Cluster Processes”, describes other options for ndbd.

When ndbd starts, it actually initiates two processes. The first of these is called the “angel process”; its only job is to discover when the execution process has been completed, and then to restart the ndbd process if it is configured to do so. Thus, if you attempt to kill ndbd via the Unix kill command, it is necessary to kill both processes, beginning with the angel process. The preferred method of terminating an ndbd process is to use the management client and stop the process from there.

The execution process uses one thread for reading, writing, and scanning data, as well as all other activities. This thread is implemented asynchronously so that it can easily handle thousands of concurrent activites. In addition, a watch-dog thread supervises the execution thread to make sure that it does not hang in an endless loop. A pool of threads handles file I/O, with each thread able to handle one open file. Threads can also be used for transporter connections by the transporters in the ndbd process. In a multi-processor system performing a large number of operations (including updates), the ndbd process can consume up to 2 CPUs if permitted to do so.

For a machine with many CPUs it is possible to use several ndbd processes which belong to different node groups; however, such a configuration is still considered experimental and is not supported for MySQL 5.1 in a production setting. See Section 15.13, “Known Limitations of MySQL Cluster”.

15.6.3. ndb_mgmd, the Management Server Process

The management server is the process that reads the cluster configuration file and distributes this information to all nodes in the cluster that request it. It also maintains a log of cluster activities. Management clients can connect to the management server and check the cluster's status.

It is not strictly necessary to specify a connectstring when starting the management server. However, if you are using more than one management server, a connectstring should be provided and each node in the cluster should specify its node ID explicitly.

See Section, “The Cluster connectstring, for information about using connectstrings. Section 15.6.5, “Command Options for MySQL Cluster Processes”, describes other options for ndb_mgmd.

The following files are created or used by ndb_mgmd in its starting directory, and are placed in the DataDir as specified in the config.ini configuration file. In the list that follows, node_id is the unique node identifier.

  • config.ini is the configuration file for the cluster as a whole. This file is created by the user and read by the management server. Section 15.4, “MySQL Cluster Configuration”, discusses how to set up this file.

  • ndb_node_id_cluster.log is the cluster events log file. Examples of such events include checkpoint startup and completion, node startup events, node failures, and levels of memory usage. A complete listing of cluster events with descriptions may be found in Section 15.7, “Management of MySQL Cluster”.

    When the size of the cluster log reaches one million bytes, the file is renamed to ndb_node_id_cluster.log.seq_id, where seq_id is the sequence number of the cluster log file. (For example: If files with the sequence numbers 1, 2, and 3 already exist, the next log file is named using the number 4.)

  • ndb_node_id_out.log is the file used for stdout and stderr when running the management server as a daemon.

  • is the process ID file used when running the management server as a daemon.

15.6.4. ndb_mgm, the Management Client Process

The management client process is actually not needed to run the cluster. Its value lies in providing a set of commands for checking the cluster's status, starting backups, and performing other administrative functions. The management client accesses the management server using a C API. Advanced users can also employ this API for programming dedicated management processes to perform tasks similar to those performed by ndb_mgm.

To start the management client, it is necessary to supply the hostname and port number of the management server:

shell> ndb_mgm [host_name [port_num]]

For example:

shell> ndb_mgm 1186

The default hostname and port number are localhost and 1186, respectively.

Additional information about using ndb_mgm can be found in Section, “Command Options for ndb_mgm, and Section 15.7.2, “Commands in the Management Client”.

15.6.5. Command Options for MySQL Cluster Processes

All MySQL Cluster executables (except for mysqld) take the options described in this section. Users of earlier MySQL Cluster versions should note that some of these options have been changed to make them consistent with one another as well as with mysqld. You can use the --help option with any MySQL Cluster executable to view a list of the options which it supports.

The following options are common to all MySQL Cluster executables:

  • --help --usage, -?

    Prints a short list with descriptions of the available command options.

  • --connect-string=connect_string, -c connect_string

    connect_string sets the connectstring to the management server as a command option.

    shell> ndbd --connect-string="nodeid=2;"
  • --debug[=options]

    This option can be used only for versions compiled with debugging enabled. It is used to enable output from debug calls in the same manner as for the mysqld process.

  • --execute=command, -e command

    Can be used to send a command to a Cluster executable from the system shell. For example, either of the following:

    shell> ndb_mgm -e "SHOW"


    shell> ndb_mgm --execute="SHOW"

    is equivalent to

    ndb_mgm> SHOW

    This is analogous to how the --execute or -e option works with the mysql command-line client. See Section 4.3.1, “Using Options on the Command Line”.

  • --version, -V

    Prints the MySQL Cluster version number of the executable. The version number is relevant because not all versions can be used together, and the MySQL Cluster startup process verifies that the versions of the binaries being used can co-exist in the same cluster. This is also important when performing an online (rolling) software upgrade or downgrade of MySQL Cluster. (See Section 15.5.1, “Performing a Rolling Restart of the Cluster”).

The next few sections describe options specific to individual NDB programs. MySQL Cluster-Related Command Options for mysqld

  • --ndb-connectstring=connect_string

    When using the NDB Cluster storage engine, this option specifies the management server that distributes cluster configuration data.

  • --ndbcluster

    The NDB Cluster storage engine is necessary for using MySQL Cluster. If a mysqld binary includes support for the NDB Cluster storage engine, the engine is disabled by default. Use the --ndbcluster option to enable it. Use --skip-ndbcluster to explicitly disable the engine. Command Options for ndbd

For options common to all NDB programs, see Section 15.6.5, “Command Options for MySQL Cluster Processes”.

  • --bind-address

    Causes ndbd to bind to a specific network interface. This option has no default value.

    This option was added in MySQL 5.1.12.

  • --daemon, -d

    Instructs ndbd to execute as a daemon process. This is the default behavior. --nodaemon can be used to prevent the process from running as a daemon.

  • --initial

    Instructs ndbd to perform an initial start. An initial start erases any files created for recovery purposes by earlier instances of ndbd. It also re-creates recovery log files. Note that on some operating systems this process can take a substantial amount of time.

    An --initial start is to be used only the very first time that the ndbd process is started because it removes all files from the Cluster filesystem and re-creates all REDO log files. The exceptions to this rule are:

    • When performing a software upgrade which has changed the contents of any files.

    • When restarting the node with a new version of ndbd.

    • As a measure of last resort when for some reason the node restart or system restart repeatedly fails. In this case, be aware that this node can no longer be used to restore data due to the destruction of the data files.

    This option does not affect either of the following:

    It is possible to achieve the same effect by deleting by other means (such as using rm -r -f) all files and directories in the data node's DataDir — with the possible exception of the BACKUP directory in DataDir, should you wish to retain any backups that have been created on that data node — and then starting ndbd without having to use the --initial option. This may be useful when scripting Cluster administrative tasks.

  • --initial-start

    This option is used when performing a partial initial start of the cluster. Each node should be started with this option, as well as --no-wait-nodes.

    For example, suppose you have a 4-node cluster whose data nodes have the IDs 2, 3, 4, and 5, and you wish to perform a partial initial start using only nodes 2, 4, and 5 — that is, omitting node 3:

    ndbd --ndbd-nodeid=2 --no-wait-nodes=3 --initial-start
    ndbd --ndbd-nodeid=4 --no-wait-nodes=3 --initial-start
    ndbd --ndbd-nodeid=5 --no-wait-nodes=3 --initial-start

    This option was added in MySQL 5.1.9.

  • --nowait-nodes=node_id_1[, node_id_2[, ...]]

    This option takes a list of data nodes which for which the cluster will not wait for before starting.

    This can be used to start the cluster in a partitioned state. For example, to start the cluster with only half of the data nodes (nodes 2, 3, 4, and 5) running in a 4-node cluster, you can start each ndbd process with --nowait-nodes=3,5. In this case, the cluster starts as soon as nodes 2 and 4 connect, and does not wait StartPartitionedTimeout milliseconds for nodes 3 and 5 to connect as it would otherwise.

    If you wanted to start up the same cluster as in the previous example without one ndbd — say, for example, that the host machine for node 3 has suffered a hardware failure — then start nodes 2, 4, and 5 with --no-wait-nodes=3. Then the cluster will start as soon as nodes 2, 4, and 5 connect and will not wait for node 3 to start.

    This option was added in MySQL 5.1.9.

  • --nodaemon

    Instructs ndbd not to start as a daemon process. This is useful when ndbd is being debugged and you want output to be redirected to the screen.

  • --nostart, -n

    Instructs ndbd not to start automatically. When this option is used, ndbd connects to the management server, obtains configuration data from it, and initializes communication objects. However, it does not actually start the execution engine until specifically requested to do so by the management server. This can be accomplished by issuing the proper START command in the management client (see Section 15.7.2, “Commands in the Management Client”). Command Options for ndb_mgmd

For options common to NDB programs, see Section 15.6.5, “Command Options for MySQL Cluster Processes”.

  • --config-file=filename, -f filename

    Instructs the management server as to which file it should use for its configuration file. This option must be specified. The filename defaults to config.ini.

    Note: This option also can be given as -c file_name, but this shortcut is obsolete and should not be used in new installations.

  • --daemon, -d

    Instructs ndb_mgmd to start as a daemon process. This is the default behavior.

  • --nodaemon

    Instructs ndb_mgmd not to start as a daemon process. Command Options for ndb_mgm

For options common to NDB programs, see Section 15.6.5, “Command Options for MySQL Cluster Processes”.

  • --try-reconnect=number

    If the connection to the management server is broken, the node tries to reconnect to it every 5 seconds until it succeeds. By using this option, it is possible to limit the number of attempts to number before giving up and reporting an error instead.

15.7. Management of MySQL Cluster

Managing a MySQL Cluster involves a number of tasks, the first of which is to configure and start MySQL Cluster. This is covered in Section 15.4, “MySQL Cluster Configuration”, and Section 15.6, “Process Management in MySQL Cluster”.

The following sections cover the management of a running MySQL Cluster.

There are essentially two methods of actively managing a running MySQL Cluster. The first of these is through the use of commands entered into the management client whereby cluster status can be checked, log levels changed, backups started and stopped, and nodes stopped and started. The second method involves studying the contents of the cluster log ndb_node_id_cluster.log; this is usually found in the management server's DataDir directory, but this location can be overridden using the LogDestination option — see Section, “Defining the Management Server”, for details. (Recall that node_id represents the unique identifier of the node whose activity is being logged.) The cluster log contains event reports generated by ndbd. It is also possible to send cluster log entries to a Unix system log.

15.7.1. MySQL Cluster Startup Phases

This section describes the steps involved when the cluster is started.

There are several different startup types and modes, as shown here:

  • Initial Start: The cluster starts with a clean filesystem on all data nodes. This occurs either when the cluster started for the very first time, or when it is restarted using the --initial option.

    Note: Disk Data files are not removed when restarting a node using --initial.

  • System Restart: The cluster starts and reads data stored in the data nodes. This occurs when the cluster has been shut down after having been in use, when it is desired for the cluster to resume operations from the point where it left off.

  • Node Restart: This is the online restart of a cluster node while the cluster itself is running.

  • Initial Node Restart: This is the same as a node restart, except that the node is reinitialized and started with a clean filesystem.

Prior to startup, each data node (ndbd process) must be initialized. Initialization consists of the following steps:

  1. Obtain a Node ID.

  2. Fetch configuration data.

  3. Allocate ports to be used for inter-node communications.

  4. Allocate memory according to settings obtained from the configuration file.

When a data node or SQL node first connects to the management node, it reserves a cluster node ID. To make sure that no other node allocates the same node ID, this ID is retained until the node has managed to connect to the cluster and at least one ndbd reports that this node is connected. This retention of the node ID is guarded by the connection between the node in question and ndb_mgmd.

Normally, in the event of a problem with the node, the node disconnects from the management server, the socket used for the connection is closed, and the reserved node ID is freed. However, if a node is disconnected abruptly — for example, due to a hardware failure in one of the cluster hosts, or because of network issues — the normal closing of the socket by the operating system may not take place. In this case, the node ID continues to be reserved and not released until a TCP timeout occurs 10 or so minutes later.

To take care of this problem, you can use PURGE STALE SESSIONS. Running this statement forces all reserved node IDs to be checked; any that are not being used by nodes actually connected to the cluster are then freed.

Beginning with MySQL 5.1.11, timeout handling of node ID assignments is implemented. This performs the ID usage checks automatically after approximately 20 seconds, so that PURGE STALE SESSIONS should no longer be necessary in a normal Cluster start.

After each data node has been initialized, the cluster startup process can proceed. The stages which the cluster goes through during this process are listed here:

  • Stage 0

    Clear the cluster filesystem. This stage occurs only if the cluster was started with the --initial option.

  • Stage 1

    This stage sets up Cluster connections, establishes inter-node communications, and starts Cluster heartbeats.

    Note: When one or more nodes hang in Phase 1 while the remaining node or nodes hang in Phase 2, this often indicates network problems. One possible cause of such issues is one or more cluster hosts having multiple network interfaces. Another common source of problems causing this condition is the blocking of TCP/IP ports needed for communications between cluster nodes. In the latter case, this is often due to a misconfigured firewall.

  • Stage 2

    The arbitrator node is elected. If this is a system restart, the cluster determines the latest restorable global checkpoint.

  • Stage 3

    This stage initializes a number of internal cluster variables.

  • Stage 4

    For an initial start or initial node restart, the redo log files are created. The number of these files is equal to NoOfFragmentLogFiles.

    For a system restart:

    • Read schema or schemas.

    • Read data from the local checkpoint and undo logs.

    • Apply all redo information until the latest restorable global checkpoint has been reached.

    For a node restart, find the tail of the redo log.

  • Stage 5

    If this is an initial start, create the SYSTAB_0 and NDB$EVENTS internal system tables.

    For a node restart or an initial node restart:

    1. The node is included in transaction handling operations.

    2. The node schema is compared with that of the master and synchronized with it.

    3. Synchronize data received in the form of INSERT from the other data nodes in this node's node group.

    4. In all cases, wait for complete local checkpoint as determined by the arbitrator.

  • Stage 6

    Update internal variables.

  • Stage 7

    Update internal variables.

  • Stage 8

    In a system restart, rebuild all indexes.

  • Stage 9

    Update internal variables.

  • Stage 10

    At this point in a node restart or initial node restart, APIs may connect to the node and begin to receive events.

  • Stage 11

    At this point in a node restart or initial node restart, event delivery is handed over to the node joining the cluster. The newly-joined node takes over responsibility for delivering its primary data to subscribers.

After this process is completed for an initial start or system restart, transaction handling is enabled. For a node restart or initial node restart, completion of the startup process means that the node may now act as a transaction coordinator.

15.7.2. Commands in the Management Client

In addition to the central configuration file, a cluster may also be controlled through a command-line interface available through the management client ndb_mgm. This is the primary administrative interface to a running cluster.

Commands for the event logs are given in Section 15.7.3, “Event Reports Generated in MySQL Cluster”; commands for creating backups and restoring from backup are provided in Section 15.8, “On-line Backup of MySQL Cluster”.

The management client has the following basic commands. In the listing that follows, node_id denotes either a database node ID or the keyword ALL, which indicates that the command should be applied to all of the cluster's data nodes.

  • HELP

    Displays information on all available commands.

  • SHOW

    Displays information on the cluster's status.

    Note: In a cluster where multiple management nodes are in use, this command displays information only for data nodes that are actually connected to the current management server.

  • node_id START

    Brings online the data node identified by node_id (or all data nodes).

    ALL START works on all data nodes only, and does not affect management nodes.

    Important: To use this command to bring a data node online, the data node must have been started using ndbd --nostart or ndbd -n.

  • node_id STOP

    Starts the data or management node identified by node_id. Note that ALL STOP works to start all data nodes only, and does not affect management nodes.

    A node affected by this command disconnects from the cluster, and its associated ndbd or ndb_mgmd process terminates.

  • node_id RESTART [-n] [-i]

    Restarts the data node identified by node_id (or all data nodes).

    Using the -i option with RESTART causes the data node to perform an initial restart; that is, the node's filesystem is deleted and recreated. The effect is the same as that obtained from stopping the data node process and then starting it again using ndbd --initial from the system shell. Note that backup files and Disk Data files are not removed when this option is used.

    Using the -n option causes the data node process to be restarted, but the data node is not actually brought online until the appropriate START command is issued. The effect of this option is the same as that obtained from stopping the data node and then starting it again using ndbd --nostart or ndbd -n from the system shell.

  • node_id STATUS

    Displays status information for the data node identified by node_id (or for all data nodes).


    Enters single user mode, whereby only the MySQL server identified by the node ID node_id is allowed to access the database.

    It is not possible in MySQL 5.1 for data nodes to join the cluster while it is running in single user mode. (See Bug#20395 for more information.)


    Exits single user mode, allowing all SQL nodes (that is, all running mysqld processes) to access the database.


    Terminates the management client.

    This command does not affect any nodes connected to the cluster.


    Shuts down all cluster data nodes and management nodes. To exit the management client after this has been done, use EXIT or QUIT.

    This command does not shut down any SQL nodes or API nodes that are connected to the cluster.

15.7.3. Event Reports Generated in MySQL Cluster

In this section, we discuss the types of event logs provided by MySQL Cluster, and the types of events that are logged.

MySQL Cluster provides two types of event log:

  • The cluster log, which includes events generated by all cluster nodes. The cluster log is the log recommended for most uses because it provides logging information for an entire cluster in a single location.

    By default, the cluster log is saved to a file named ndb_node_id_cluster.log, (where node_id is the node ID of the management server) in the same directory where the ndb_mgm binary resides.

    Cluster logging information can also be sent to stdout or a syslog facility in addition to or instead of being saved to a file, as determined by the values set for the DataDir and LogDestination configuration parameters. See Section, “Defining the Management Server”, for more information about these parameters.

  • Node logs are local to each node.

    Output generated by node event logging is written to the file ndb_node_id_out.log (where node_id is the node's node ID) in the node's DataDir. Node event logs are generated for both management nodes and data nodes.

    Node logs are intended to be used only during application development, or for debugging application code.

Both types of event logs can be set to log different subsets of events.

Each reportable event can be distinguished according to three different criteria:

  • Category: This can be any one of the following values: STARTUP, SHUTDOWN, STATISTICS, CHECKPOINT, NODERESTART, CONNECTION, ERROR, or INFO.

  • Priority: This is represented by one of the numbers from 1 to 15 inclusive, where 1 indicates “most important” and 15 “least important.

  • Severity Level: This can be any one of the following values: ALERT, CRITICAL, ERROR, WARNING, INFO, or DEBUG.

Both the cluster log and the node log can be filtered on these properties.

The format used in the cluster log is as shown here:

2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 1: Data usage is 2%(60 32K pages of total 2560)
2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 1: Index usage is 1%(24 8K pages of total 2336)
2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 1: Resource 0 min: 0 max: 639 curr: 0
2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 2: Data usage is 2%(76 32K pages of total 2560)
2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 2: Index usage is 1%(24 8K pages of total 2336)
2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 2: Resource 0 min: 0 max: 639 curr: 0
2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 3: Data usage is 2%(58 32K pages of total 2560)
2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 3: Index usage is 1%(25 8K pages of total 2336)
2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 3: Resource 0 min: 0 max: 639 curr: 0
2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 4: Data usage is 2%(74 32K pages of total 2560)
2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 4: Index usage is 1%(25 8K pages of total 2336)
2007-01-26 19:35:55 [MgmSrvr] INFO     -- Node 4: Resource 0 min: 0 max: 639 curr: 0
2007-01-26 19:39:42 [MgmSrvr] INFO     -- Node 4: Node 9 Connected
2007-01-26 19:39:42 [MgmSrvr] INFO     -- Node 1: Node 9 Connected
2007-01-26 19:39:42 [MgmSrvr] INFO     -- Node 1: Node 9: API version 5.1.15
2007-01-26 19:39:42 [MgmSrvr] INFO     -- Node 2: Node 9 Connected
2007-01-26 19:39:42 [MgmSrvr] INFO     -- Node 2: Node 9: API version 5.1.15
2007-01-26 19:39:42 [MgmSrvr] INFO     -- Node 3: Node 9 Connected
2007-01-26 19:39:42 [MgmSrvr] INFO     -- Node 3: Node 9: API version 5.1.15
2007-01-26 19:39:42 [MgmSrvr] INFO     -- Node 4: Node 9: API version 5.1.15
2007-01-26 19:59:22 [MgmSrvr] ALERT    -- Node 2: Node 7 Disconnected
2007-01-26 19:59:22 [MgmSrvr] ALERT    -- Node 2: Node 7 Disconnected

Each line in the cluster log contains the following information:

  • A timestamp in YYYY-MM-DD HH:MM:SS format.

  • The type of node which is performing the logging. In the cluster log, this is always [MgmSrvr].

  • The severity of the event.

  • The ID of the node reporting the event.

  • A description of the event. The most common types of events to appear in the log are connections and disconnections between different nodes in the cluster, and when checkpoints occur. In some cases, the description may contain status information. Logging Management Commands

The following management commands are related to the cluster log:


    Turns the cluster log on.


    Turns the cluster log off.


    Provides information about cluster log settings.

  • node_id CLUSTERLOG category=threshold

    Logs category events with priority less than or equal to threshold in the cluster log.

  • CLUSTERLOG FILTER severity_level

    Toggles cluster logging of events of the specified severity_level.

The following table describes the default setting (for all data nodes) of the cluster log category threshold. If an event has a priority with a value lower than or equal to the priority threshold, it is reported in the cluster log.

Note that events are reported per data node, and that the threshold can be set to different values on different nodes.

CategoryDefault threshold (All data nodes)

The STATISTICS category can provide a great deal of useful data. See Section, “Using CLUSTERLOG STATISTICS, for more information.

Thresholds are used to filter events within each category. For example, a STARTUP event with a priority of 3 is not logged unless the threshold for STARTUP is changed to 3 or lower. Only events with priority 3 or lower are sent if the threshold is 3.

The following table shows the event severity levels. (Note: These correspond to Unix syslog levels, except for LOG_EMERG and LOG_NOTICE, which are not used or mapped.)

1ALERTA condition that should be corrected immediately, such as a corrupted system database
2CRITICALCritical conditions, such as device errors or insufficient resources
3ERRORConditions that should be corrected, such as configuration errors
4WARNINGConditions that are not errors, but that might require special handling
5INFOInformational messages
6DEBUGDebugging messages used for NDB Cluster development

Event severity levels can be turned on or off (using CLUSTERLOG FILTER — see above). If a severity level is turned on, then all events with a priority less than or equal to the category thresholds are logged. If the severity level is turned off then no events belonging to that severity level are logged. Log Events

An event report reported in the event logs has the following format:

datetime [string] severity -- message

For example:

09:19:30 2005-07-24 [NDB] INFO -- Node 4 Start phase 4 completed

This section discusses all reportable events, ordered by category and severity level within each category.

In the event descriptions, GCP and LCP mean “Global Checkpoint” and “Local Checkpoint,” respectively.


These events are associated with connections between Cluster nodes.

EventPrioritySeverity LevelDescription
data nodes connected8INFOData nodes connected
data nodes disconnected8INFOData nodes disconnected
Communication closed8INFOSQL node or data node connection closed
Communication opened8INFOSQL node or data node connection opened


The logging messages shown here are associated with checkpoints.

EventPrioritySeverity LevelDescription
LCP stopped in calc keep GCI0ALERTLCP stopped
Local checkpoint fragment completed11INFOLCP on a fragment has been completed
Global checkpoint completed10INFOGCP finished
Global checkpoint started9INFOStart of GCP: REDO log is written to disk
Local checkpoint completed8INFOLCP completed normally
Local checkpoint started7INFOStart of LCP: data written to disk
Report undo log blocked7INFOUNDO logging blocked; buffer near overflow


The following events are generated in response to the startup of a node or of the cluster and of its success or failure. They also provide information relating to the progress of the startup process, including information concerning logging activities.

EventPrioritySeverity LevelDescription
Internal start signal received STTORRY15INFOBlocks received after completion of restart
Undo records executed15INFO 
New REDO log started10INFOGCI keep X, newest restorable GCI Y
New log started10INFOLog part X, start MB Y, stop MB Z
Node has been refused for inclusion in the cluster8INFONode cannot be included in cluster due to misconfiguration, inability to establish communication, or other problem
data node neighbors8INFOShows neighboring data nodes
data node start phase X completed4INFOA data node start phase has been completed
Node has been successfully included into the cluster3INFODisplays the node, managing node, and dynamic ID
data node start phases initiated1INFONDB Cluster nodes starting
data node all start phases completed1INFONDB Cluster nodes started
data node shutdown initiated1INFOShutdown of data node has commenced
data node shutdown aborted1INFOUnable to shut down data node normally


The following events are generated when restarting a node and relate to the success or failure of the node restart process.

EventPrioritySeverity LevelDescription
Node failure phase completed8ALERTReports completion of node failure phases
Node has failed, node state was X8ALERTReports that a node has failed
Report arbitrator results2ALERTThere are eight different possible results for arbitration attempts:
  • Arbitration check failed — less than 1/2 nodes left

  • Arbitration check succeeded — node group majority

  • Arbitration check failed — missing node group

  • Network partitioning — arbitration required

  • Arbitration succeeded — affirmative response from node X

  • Arbitration failed - negative response from node X

  • Network partitioning - no arbitrator available

  • Network partitioning - no arbitrator configured

Completed copying a fragment10INFO 
Completed copying of dictionary information8INFO 
Completed copying distribution information8INFO 
Starting to copy fragments8INFO 
Completed copying all fragments8INFO 
GCP takeover started7INFO 
GCP takeover completed7INFO 
LCP takeover started7INFO 
LCP takeover completed (state = X)7INFO 
Report whether an arbitrator is found or not6INFOThere are seven different possible outcomes when seeking an arbitrator:
  • Management server restarts arbitration thread [state=X]

  • Prepare arbitrator node X [ticket=Y]

  • Receive arbitrator node X [ticket=Y]

  • Started arbitrator node X [ticket=Y]

  • Lost arbitrator node X - process failure [state=Y]

  • Lost arbitrator node X - process exit [state=Y]

  • Lost arbitrator node X <error msg> [state=Y]


The following events are of a statistical nature. They provide information such as numbers of transactions and other operations, amount of data sent or received by individual nodes, and memory usage.

EventPrioritySeverity LevelDescription
Report job scheduling statistics9INFOMean internal job scheduling statistics
Sent number of bytes9INFOMean number of bytes sent to node X
Received # of bytes9INFOMean number of bytes received from node X
Report transaction statistics8INFONumbers of: transactions, commits, reads, simple reads, writes, concurrent operations, attribute information, and aborts
Report operations8INFONumber of operations
Report table create7INFO 
Memory usage5INFOData and index memory usage (80%, 90%, and 100%)

ERROR Events

These events relate to Cluster errors and warnings. The presence of one or more of these generally indicates that a major malfunction or failure has occurred.

Dead due to missed heartbeat8ALERTNode X declared “dead” due to missed heartbeat
Transporter errors2ERROR 
Transporter warnings8WARNING 
Missed heartbeats8WARNINGNode X missed heartbeat #Y
General warning events2WARNING 

INFO Events

These events provide general information about the state of the cluster and activities associated with Cluster maintenance, such as logging and heartbeat transmission.

Sent heartbeat12INFOHeartbeat sent to node X
Create log bytes11INFOLog part, log file, MB
General information events2INFO Using CLUSTERLOG STATISTICS

The NDB management client's CLUSTERLOG STATISTICS command can provide a number of useful statistics in its output. The following statistics are reported by the transaction coordinator:

StatisticDescription (Number of...)
Trans. CountTransactions attempted with this node as coordinator
Commit CountTransactions committed with this node as coordinator
Read CountPrimary key reads (all)
Simple Read CountPrimary key reads reading the latest committed value
Write CountPrimary key writes (includes all INSERT, UPDATE, and DELETE operations)
AttrInfoCountData words used to describe all reads and writes received
Concurrent OperationsAll concurrent operations ongoing at the moment the report is taken
Abort CountTransactions with this node as coordinator that were aborted
ScansScans (all)
Range ScansIndex scans

The ndbd process has a scheduler that runs in an infinite loop. During each loop scheduler performs the following tasks:

  1. Read any incoming messages from sockets into a job buffer.

  2. Check whether there are any timed messages to be executed; if so, put these into the job buffer as well.

  3. Execute (in a loop) any messages in the job buffer.

  4. Send any distributed messages that were generated by executing the messages in the job buffer.

  5. Wait for any new incoming messages.

The number of loops executed in the third step is reported as the Mean Loop Counter. This statistic increases in size as the utilisation of the TCP/IP buffer improves. You can use this to monitor performance as you add new processes to the cluster.

The Mean send size and Mean receive size statistics allow you to gauge the efficiency of writes and reads (respectively) between nodes. These values are given in bytes. Higher values mean a lower cost per byte sent or received; the maximum is 64k.

To cause all cluster log statistics to be logged, you can use the following command in the NDB management client:


Note: Setting the threshold for STATISTICS to 15 causes the cluster log to become very verbose, and to gow quite rapidly in size, in direct proportion to the number of cluster nodes and the amount of activity on the cluster.

15.7.4. Single User Mode

Single user mode allows the database administrator to restrict access to the database system to a single API node, such as a MySQL server (SQL node) or an instance of ndb_restore. When entering single user mode, connections to all other API nodes are closed gracefully and all running transactions are aborted. No new transactions are permitted to start.

Once the cluster has entered single user mode, only the designated API node is granted access to the database.

You can use the ALL STATUS command to see when the cluster has entered single user mode.



After this command has executed and the cluster has entered single user mode, the API node whose node ID is 5 becomes the cluster's only permitted user.

The node specified in the preceding command must be an API node; attempting to specify any other type of node will be rejected.

Note: When the preceding commmand is invoked, all transactions running on the designated node are aborted, the connection is closed, and the server must be restarted.

The command EXIT SINGLE USER MODE changes the state of the cluster's data nodes from single user mode to normal mode. API nodes — such as MySQL Servers — waiting for a connection (that is, waiting for the cluster to become ready and available), are again permitted to connect. The API node denoted as the single-user node continues to run (if still connected) during and after the state change.



There are two recommended ways to handle a node failure when running in single user mode:

  • Method 1:

    1. Finish all single user mode transactions

    2. Issue the EXIT SINGLE USER MODE command

    3. Restart the cluster's data nodes

  • Method 2:

    Restart database nodes prior to entering single user mode.

15.8. On-line Backup of MySQL Cluster

This section describes how to create a backup and how to restore the database from a backup at a later time.

15.8.1. Cluster Backup Concepts

A backup is a snapshot of the database at a given time. The backup consists of three main parts:

  • Metadata: the names and definitions of all database tables

  • Table records: the data actually stored in the database tables at the time that the backup was made

  • Transaction log: a sequential record telling how and when data was stored in the database

Each of these parts is saved on all nodes participating in the backup. During backup, each node saves these three parts into three files on disk:

  • BACKUP-backup_id.node_id.ctl

    A control file containing control information and metadata. Each node saves the same table definitions (for all tables in the cluster) to its own version of this file.


    A data file containing the table records, which are saved on a per-fragment basis. That is, different nodes save different fragments during the backup. The file saved by each node starts with a header that states the tables to which the records belong. Following the list of records there is a footer containing a checksum for all records.

  • BACKUP-backup_id.node_id.log

    A log file containing records of committed transactions. Only transactions on tables stored in the backup are stored in the log. Nodes involved in the backup save different records because different nodes host different database fragments.

In the listing above, backup_id stands for the backup identifier and node_id is the unique identifier for the node creating the file.

15.8.2. Using The Management Client to Create a Backup

Before starting a backup, make sure that the cluster is properly configured for performing one. (See Section 15.8.4, “Configuration for Cluster Backup”.)

Creating a backup using the management client involves the following steps:

  1. Start the management client (ndb_mgm).

  2. Execute the command START BACKUP.

  3. The management client responds as shown here:

    Waiting for completed, this may take several minutes
    Node 1: Backup backup_id started from node management_node_id

    Here, backup_id is the unique identifier for this particular backup. (This identifier will also be saved in the cluster log, if it has not been configured otherwise.) management_node_id is the node ID of the management to which the management client is connected.

    This means that the cluster has received and processed the backup request. It does not mean that the backup has been completed.

    Note: Backup messages were not recorded in the cluster log in MySQL 5.1.12 or 5.1.13. The logging of backup operations was restored in MySQL 5.1.14 (see Bug#24544).

  4. When the backup is completed, the management client will indicate this as shown here:

    Node 1: Backup backup_id started from node management_node_id completed
     StartGCP: 417599 StopGCP: 417602
     #Records: 105957 #LogRecords: 0
     Data: 99719356 bytes Log: 0 bytes

    The values shown for StartGCP, StopGCP, #Records, #LogRecords, Data, and Log will vary according to the specifics of your cluster.

Cluster backups are created by default in the BACKUP subdirectory of the DataDir on each data node. This can be overridden for one or more data nodes individually, or for all cluster data nodes in the config.ini file using the BackupDataDir configuration parameter as discussed in Identifying Data Nodes. The backup files created for a backup with a given backup_id are stored in a subdirectory named BACKUP-backup_id in the backup directory.

To abort a backup that is already in progress:

  1. Start the management client.

  2. Execute this command:

    ndb_mgm> ABORT BACKUP backup_id

    The number backup_id is the identifier of the backup that was included in the response of the management client when the backup was started (in the message Backup backup_id started from node management_node_id).

  3. The management client will acknowledge the abort request with Abort of backup backup_id ordered. Note: At this point, the management client has not yet received a response from the cluster data nodes to this request, and the backup has not yet actually been aborted.

  4. After the backup has been aborted, the management client will report this fact in a manner similar to what is shown here:

    Node 1: Backup 3 started from 5 has been aborted. Error: 1321 - Backup aborted by user request: Permanent error: User defined error
    Node 3: Backup 3 started from 5 has been aborted. Error: 1323 - 1323: Permanent error: Internal error
    Node 2: Backup 3 started from 5 has been aborted. Error: 1323 - 1323: Permanent error: Internal error
    Node 4: Backup 3 started from 5 has been aborted. Error: 1323 - 1323: Permanent error: Internal error

    In this example, we have shown sample output for a cluster with 4 data nodes, where the sequence number of the backup to be aborted is 3, and the management node to which the cluster management client is connected has the node ID 5. The first node to complete its part in aborting the backup reports that the reason for the abort was due to a request by the user. (The remaining nodes report that the backup was aborted due to an unspecified internal error.) Note: There is no guarantee that the cluster nodes will respond to an ABORT BACKUP command in any particular order.

    The Backup backup_id started from node management_node_id has been aborted messages mean that the backup has been terminated and that all files relating to this backup have been removed from the cluster filesystem.

It is also possible to abort a backup in progress from a system shell using this command:

shell> ndb_mgm -e "ABORT BACKUP backup_id"

Note: If there is no backup with ID backup_id running when an ABORT BACKUP is issued, the management client makes no response, nor is it indicated in the cluster log that an invalid abort command was sent.

15.8.3. How to Restore a Cluster Backup

The cluster restoration program is implemented as a separate command-line utility ndb_restore, which can normally be found in the MySQL bin directory. This program reads the files created as a result of the backup and inserts the stored information into the database.

ndb_restore must be executed once for each of the backup files that were created by the START BACKUP command used to create the backup (see Section 15.8.2, “Using The Management Client to Create a Backup”). This is equal to the number of data nodes in the cluster at the time that the backup was created.

Note: Before using ndb_restore, it is recommended that the cluster be running in single user mode, unless you are restoring multiple data nodes in parallel. See Section 15.7.4, “Single User Mode”, for more information about single user mode.

Typical options for this utility are shown here:

ndb_restore [-c connectstring] -n node_id [-m] -b backup_id -r /path/to/backup/files

The -c option is used to specify a connectstring which tells ndb_restore where to locate the cluster management server. (See Section, “The Cluster connectstring, for information on connectstrings.) If this option is not used, then ndb_restore attempts to connect to a management server on localhost:1186. This utility acts as a cluster API node, and so requires a free connection “slot” to connect to the cluster management server. This means that there must be at least one [API] or [MYSQLD] section that can be used by it in the cluster config.ini file. It is a good idea to keep at least one empty [API] or [MYSQLD] section in config.ini that is not being used for a MySQL server or other application for this reason (see Section, “Defining SQL and Other API Nodes”).

You can verify that ndb_restore is connected to the cluster by using the SHOW command in the ndb_mgm management client. You can also accomplish this from a system shell, as shown here:

shell> ndb_mgm -e "SHOW"

-n is used to specify the node ID of the data node on which the backups were taken.

The first time you run the ndb_restore restoration program, you also need to restore the metadata. In other words, you must re-create the database tables — this can be done by running it with the -m option. Note that the cluster should have an empty database when starting to restore a backup. (In other words, you should start ndbd with --initial prior to performing the restore. You should also remove manually any Disk Data files present in the data node's DataDir.)

The -b option is used to specify the ID or sequence number of the backup, and is the same number shown by the management client in the Backup backup_id completed message displayed upon completion of a backup. (See Section 15.8.2, “Using The Management Client to Create a Backup”.)

The -r option is required, and is used to tell ndb_restore the directory where it can find the backup files. Important: When restoring cluster backups, you must be sure to restore all data nodes from backups having the same backup ID.

It is possible to restore a backup to a database with a different configuration than it was created from. For example, suppose that a backup with backup ID 12, created in a cluster with two database nodes having the node IDs 2 and 3, is to be restored to a cluster with four nodes. Then ndb_restore must be run twice — once for each database node in the cluster where the backup was taken. However, ndb_restore cannot always restore backups made from a cluster running one version of MySQL to a cluster running a different MySQL version. See Section 15.5.2, “Cluster Upgrade and Downgrade Compatibility”, for more information.

Note: For more rapid restoration, the data may be restored in parallel, provided that there is a sufficient number of cluster connections available. That is, when restoring to multiple nodes in parallel, you must have an [API] or [MYSQLD] section in the cluster config.ini file available for each concurrent ndb_restore process. However, the data files must always be applied before the logs.

A complete listing of options available for this program is shown in the following table:

Long FormShort FormDescriptionDefault Value
--backup-id-bBackup sequence ID0
--backup_pathNonePath to backup files./
--character-sets-dirNoneSpecify the directory where character set information can be foundNone
--connect, --connectstring, or --ndb-connectstring-c or -CSet the connectstring in [nodeid=node_id;][host=]host[:port] formatNone
--core-fileNoneWrite a core file in the event of an errorTRUE
--debug-#Output debug logd:t:O,/tmp/ndb_restore.trace
--dont_ignore_systab_0-fDo not ignore system table during restore — EXPERIMENTAL; not for production useFALSE
--help or --usage-?Display help message with available options and current values, then exit[N/A]
--ndb-mgmd-hostNoneSet the host and port in host[:port] format for the management server to connect to; this is the same as --connect, --connectstring, or --ndb-connectstring, but without a way to specify the nodeidNone
--ndb-nodegroup-map-zSpecifies a nodegroup map — Syntax: list of (source_nodegroup, destination_nodegroup)None
--ndb-nodeidNoneSpecify a node ID for the ndb_restore process0
--ndb-optimized-node-selectionNoneOptimize selection of nodes for transactionsTRUE
--ndb-shmNoneUse shared memory connections when availableFALSE
--no-restore-disk-objects-dDo not restore Disk Data objects such as tablespaces and log file groupsFALSE (in other words, restore Disk Data objects unless this option is used)
--nodeid-nUse backup files from node with the specified ID0
--parallelism-pSet from 1 to 1024 parallel transactions to be used during the restoration process128
--printNonePrint data and log to stdoutFALSE
--print_dataNonePrint data to stdoutFALSE
--print_logNonePrint log to stdoutFALSE
--print_metaNonePrint metadata to stdoutFALSE
--restore_data-rRestore data and logsFALSE
--restore_epoch-eRestore epoch data into the status table; the row in the cluster.apply_status having the id 0 is inserted or updated as appropriate — this is convenient when starting up replication on a MySQL Cluster replication slaveFALSE
--restore_meta-mRestore table metadataFALSE
--version-VOutput version information and exit[N/A]

Note: ndb_restore reports both temporary and permanent errors. In the case of temporary errors, it may able to recover from them. Beginning with MySQL 5.1.12, it reports Restore successful, but encountered temporary error, please look at configuration in such cases.

15.8.4. Configuration for Cluster Backup

Four configuration parameters are essential for backup:

  • BackupDataBufferSize

    The amount of memory used to buffer data before it is written to disk.

  • BackupLogBufferSize

    The amount of memory used to buffer log records before these are written to disk.

  • BackupMemory

    The total memory allocated in a database node for backups. This should be the sum of the memory allocated for the backup data buffer and the backup log buffer.

  • BackupWriteSize

    The default size of blocks written to disk. This applies for both the backup data buffer and the backup log buffer.

  • BackupMaxWriteSize

    The maximum size of blocks written to disk. This applies for both the backup data buffer and the backup log buffer.

More detailed information about these parameters can be found in Backup Parameters.

15.8.5. Backup Troubleshooting

If an error code is returned when issuing a backup request, the most likely cause is insufficient memory or disk space. You should check that there is enough memory allocated for the backup. Important: If you have set BackupDataBufferSize and BackupLogBufferSize and their sum is greater than 4MB, then you must also set BackupMemory as well. See BackupMemory.

You should also make sure that there is sufficient space on the hard drive partition of the backup target.

NDB does not support repeatable reads, which can cause problems with the restoration process. Although the backup process is “hot”, restoring a MySQL Cluster from backup is not a 100% “hot” process. This is due to the fact that, for the duration of the restore process, running transactions get non-repeatable reads from the restored data. This means that the state of the data is inconsistent while the restore is in progress.

15.9. Cluster Utility Programs

This section discusses the MySQL Cluster utility programs that can be found in the mysql/bin directory. Each of these — except for and ndb_error_reporter — is a standalone binary that can be used from a system shell, and that does not need to connect to a MySQL server (nor even requires that a MySQL server be connected to the cluster).

These utilities can also serve as examples for writing your own applications using the NDB API. The source code for most of these programs may be found in the storage/ndb/tools directory of the MySQL 5.1 tree (see Section 2.9, “MySQL Installation Using a Source Distribution”). The NDB API is not covered in this manual; please refer to the NDB API Guide for information about this API.

All of the NDB utilities are listed here with brief descriptions:

  • ndb_config: Retrieves Cluster configuration option values.

  • ndb_delete_all: Deletes all rows from a given table.

  • ndb_desc: Lists all properties of an NDB table.

  • ndb_drop_index: Drops the specified index from an NDB table.

  • ndb_drop_table: Drops an NDB table.

  • ndb_error_reporter: Can be used to gather information useful for diagnosing problems with the cluster.

  • ndb_mgm: This is the MySQL Cluster management client, which is discussed in Section 15.7.2, “Commands in the Management Client”.

  • ndb_print_backup_file: Prints diagnostic information obtained from cluster backup files.

  • ndb_print_schema_file: Prints diagnostic information obtained from cluster schema files.

  • ndb_print_sys_file: Prints diagnostic information obtained from cluster system files.

  • ndb_restore: This utility is used to restore a cluster from backup. See Section 15.8.3, “How to Restore a Cluster Backup”, for more information.

  • ndb_select_all: Prints all rows from an NDB table.

  • ndb_select_count: Gets the number of rows in one or more NDB tables.

  • ndb_show_tables: Shows all NDB tables anywhere in the cluster.

  • Examines all the tables in a given non-Cluster database and calculates the amount of storage each would require if it were converted to use the NDB storage engine.

  • ndb_waiter: Reports on the status of cluster data nodes in a manner similar to that of the management client command ALL STATUS.

Most of these utilities need to connect to a Cluster management server in order to function. The exceptions are (see below), and the following utilities which access a cluster data node filesystem and so need to be run on a data node host:

  • ndb_print_backup_file

  • ndb_print_schema_file

  • ndb_print_sys_file is a Perl script which is also intended to be used from the shell; however it is a MySQL application and must be able to connect to a MySQL server. See Section 15.9.13, “ — NDBCluster Size Requirement Estimator”, for additional requirements for using this script.

ndb_error_reporter is also a Perl script. It is used to gather cluster data node and management node logs together into a tarball to submit along with a bug report. It can use ssh or scp to access the node filesystems remotely.

Additional information about each of these utilities (except for ndb_mgm and ndb_restore) can be found in the sections that follow.

Note: All of these utilities (except for and ndb_config) can use the options discussed in Section 15.6.5, “Command Options for MySQL Cluster Processes”. Additional options specific to each utility program are discussed in the individual program listings.

The order in which these options are used is generally not important. For example, all of these commands produce exactly the same output:

  • ndb_desc -c localhost fish -d test

  • ndb_desc fish -c localhost -d test

  • ndb_desc -d test fish -c localhost

15.9.1. ndb_config

Description: This tool extracts configuration information for data nodes, SQL nodes, and API nodes from a cluster management node (and possibly its config.ini file).


ndb_config options

The options available for this utility differ somewhat from those used with the other utilities, and so are listed in their entireity in the next section, followed by some examples.


  • --usage, --help, or -?

    Causes ndb_config to print a list of available options, and then exit.

  • --version, -V

    Causes ndb_config to print a version information string, and then exit.

  • --ndb-connectstring=connect_string

    Specifies the connectstring to use in connecting to the management server. The format for the connectstring is the same as described in Section, “The Cluster connectstring, and defaults to localhost:1186.

    The use of -c as a short version for this option is supported for ndb_config beginning with MySQL 5.1.12.

  • --config-file=path-to-file

    Gives the path to the management server's configuration file (config.ini). This may be a relative or absolute path. If the management node resides on a different host from the one on which ndb_config is invoked, then an absolute path must be used.

  • --query=query-options, -q query-options

    This is a comma-delimited list of query options — that is, a list of one or more node attributes to be returned. These include id (node ID), type (node type — that is, ndbd, mysqld, or ndb_mgmd), and any configuration parameters whose values are to be obtained.

    For example, --query=id,type,indexmemory,datamemory would return the node ID, node type, DataMemory, and IndexMemory for each node.

    Note: If a given parameter is not applicable to a certain type of node, than an empty string is returned for the corresponding value. See the examples later in this section for more information.

  • --host=hostname

    Specifies the hostname of the node for which configuration information is to be obtained.

  • --id=node_id, --nodeid=node_id

    Used to specify the node ID of the node for which configuration information is to be obtained.

  • --nodes

    (Tells ndb_config to print information from parameters defined in [ndbd] sections only. Currently, using this option has no affect, since these are the only values checked, but it may become possible in future to query parameters set in [tcp] and other sections of cluster configuration files.)

  • --type=node_type

    Filters results so that only configuration values applying to nodes of the specified node_type (ndbd, mysqld, or ndb_mgmd) are returned.

  • --fields=delimiter, -f delimiter

    Specifies a delimiter string used to separate the fields in the result. The default is “,” (the comma character).

    Note: If the delimiter contains spaces or escapes (such as \n for the linefeed character), then it must be quoted.

  • --rows=separator, -r separator

    Specifies a separator string used to separate the rows in the result. The default is a space character.

    Note: If the separator contains spaces or escapes (such as \n for the linefeed character), then it must be quoted.


  1. To obtain the node ID and type of each node in the cluster:

    shell> ./ndb_config --query=id,type --fields=':' --rows='\n'

    In this example, we used the --fields options to separate the ID and type of each node with a colon character (:), and the --rows options to place the values for each node on a new line in the output.

  2. To produce a connectstring that can be used by data, SQL, and API nodes to connect to the management server:

    shell> ./ndb_config --config-file=usr/local/mysql/cluster-data/config.ini --query=hostname,portnumber --fields=: --rows=, --type=ndb_mgmd
  3. This invocation of ndb_config checks only data nodes (using the --type option), and shows the values for each node's ID and hostname, and its DataMemory, IndexMemory, and DataDir parameters:

    shell> ./ndb_config --type=ndbd --query=id,host,datamemory,indexmemory,datadir -f ' : ' -r '\n'
    1 : : 83886080 : 18874368 : /usr/local/mysql/cluster-data
    2 : : 83886080 : 18874368 : /usr/local/mysql/cluster-data
    3 : : 83886080 : 18874368 : /usr/local/mysql/cluster-data
    4 : : 83886080 : 18874368 : /usr/local/mysql/cluster-data

    In this example, we used the short options -f and -r for setting the field delimiter and row separator, respectively.

  4. To exclude results from any host except one in particular, use the --host option:

    shell> ./ndb_config --host= -f : -r '\n' -q id,type

    In this example, we also used the short form -q to determine the attributes to be queried.

    Similarly, you can limit results to a node with a specific ID using the --id or --nodeid option.

15.9.2. ndb_delete_all

Description: Deletes all rows from the given NDB table. In some cases, this can be much faster than DELETE or even TRUNCATE.


ndb_delete_all -c connect_string tbl_name -d db_name

This deletes all rows from the table named tbl_name in the database named db_name. It is exactly equivalent to executing TRUNCATE db_name.tbl_name in MySQL.

Additional Options:

  • -t, --transactional

    Use of this option causes the delete operation to be performed as a single transaction.

    Warning: With very large tables, this using this option may cause the number of operations available to the cluster to be exceeded.

15.9.3. ndb_desc

Description: Provides a detailed description of one or more NDB tables.


ndb_desc -c connect_string tbl_name -d db_name

Sample Output:

MySQL table creation and population statements:

USE test;

    name VARCHAR(20),

    PRIMARY KEY pk (id),
    UNIQUE KEY uk (name)

    ('','guppy'), ('','tuna'), ('','shark'), 
    ('','manta ray'), ('','grouper'), ('','puffer');

Output from ndb_desc:

shell> ./ndb_desc -c localhost fish -d test -p
-- fish --
Version: 16777221
Fragment type: 5
K Value: 6
Min load factor: 78
Max load factor: 80
Temporary table: no
Number of attributes: 2
Number of primary keys: 1
Length of frm data: 268
Row Checksum: 1
Row GCI: 1
TableStatus: Retrieved
-- Attributes --
name Varchar(20;latin1_swedish_ci) NULL AT=SHORT_VAR ST=MEMORY

-- Indexes --
PRIMARY KEY(id) - UniqueHashIndex
uk(name) - OrderedIndex
PRIMARY(id) - OrderedIndex
uk$unique(name) - UniqueHashIndex

-- Per partition info --
Partition  Row count  Commit count  Frag fixed memory  Frag varsized memory
2          2          2             65536              327680
1          2          2             65536              327680
3          2          2             65536              327680

NDBT_ProgramExit: 0 - OK

Additional Options:

  • -p, --extra-partition-info

    Prints additional information about the table's partitions.

  • Information about multiple tables can be obtained in a single invocation of ndb_desc by using their names, separated by spaces. All of the tables must be in the same database.

15.9.4. ndb_drop_index

Description: Drops the specified index from an NDB table. It is recommended that you use this utility only as an example for writing NDB API applications — see the Warning later in this section for details.


ndb_drop_index -c connect_string table_name index -d db_name

The statement shown above drops the index named index from the table in the database.

Additional Options: None that are specific to this application.

Warning: Operations performed on Cluster table indexes using the NDB API are not visible to MySQL and make the table unusable by a MySQL server. If you use this program to drop an index, then try to access the table from an SQL node, an error results, as shown here:

shell> ./ndb_drop_index -c localhost dogs ix -d ctest1
Dropping index dogs/idx...OK

NDBT_ProgramExit: 0 - OK

shell> ./mysql -u jon -p ctest1
Enter password: *******
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A

Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 7 to server version: 5.1.12-beta-20060817

Type 'help;' or '\h' for help. Type '\c' to clear the buffer.

| Tables_in_ctest1 |
| a                |
| bt1              |
| bt2              |
| dogs             |
| employees        |
| fish             |
6 rows in set (0.00 sec)

mysql> SELECT * FROM dogs;
ERROR 1296 (HY000): Got error 4243 'Index not found' from NDBCLUSTER

In such a case, your only option for making the table available to MySQL again is to drop the table and re-create it. You can use either the SQL statementDROP TABLE or the ndb_drop_table utility (see Section 15.9.5, “ndb_drop_table) to drop the table.

15.9.5. ndb_drop_table

Description: Drops the specified NDB table. (If you try to use this on a table created with a storage engine other than NDB, it fails with the error 723: No such table exists.) This operation is extremely fast — in some cases, it can be an order of magnitude faster than using DROP TABLE on an NDB table from MySQL.


ndb_drop_table -c connect_string tbl_name -d db_name

Additional Options: None.

15.9.6. ndb_error_reporter

Description: Creates an archive from data node and management node log files that can be used to help diagnose bugs or other problems with a cluster. It is highly recommended that you make use of this utility when filing reports of bugs in MySQL Cluster.


ndb_error_reporter path/to/config-file [username] [--fs]

This utility is intended for use on a management node host, and requires the path to the management host configuration file (config.ini). Optionally, you can supply the name of a user that is able to access the cluster's data nodes via SSH, in order to copy the data node log files. ndb_error_reporter then includes all of these files in archive that is created in the same directory in which it is run. The archive is named ndb_error_report_YYYYMMDDHHMMSS.tar.bz2, where YYYYMMDDHHMMSS is a datetime string.

If the --fs is used, then the data node filesystems are also copied to the management host and included in the archive that is produced by this script. As data node filesystems can be extremely large even after being compressed, we ask that you please do not send archives created using this option to MySQL AB unless you are specifically requested to do so.

15.9.7. ndb_print_backup_file

Description: Obtains diagnostic information from a cluster backup file.


ndb_print_backup_file file_name

file_name is the name of a cluster backup file. This can be any of the files (.Data, .ctl, or .log file) found in a cluster backup directory. These files are found in the data node's backup directory under the subdirectory BACKUP-#, where # is the sequence number for the backup. For more information about cluster backup files and their contents, see Section 15.8.1, “Cluster Backup Concepts”.

Like ndb_print_schema_file and ndb_print_sys_file (and unlike most of the other NDB utilities that are intended to be run on a management server host or to connect to a management server) ndb_print_backup_file must be run on a cluster data node, since it accesses the data node filesystem directly. Because it does not make use of the management server, this utility can be used when the management server is not running, and even when the cluster has been completely shut down.

Additional Options: None.

15.9.8. ndb_print_schema_file

Description: Obtains diagnostic information from a cluster schema file.


ndb_print_schema_file file_name

file_name is the name of a cluster schema file.

Like ndb_print_backup_file and ndb_print_sys_file (and unlike most of the other NDB utilities that are intended to be run on a management server host or to connect to a management server) ndb_schema_backup_file must be run on a cluster data node, since it accesses the data node filesystem directly. Because it does not make use of the management server, this utility can be used when the management server is not running, and even when the cluster has been completely shut down.

Additional Options: None.

15.9.9. ndb_print_sys_file

Description: Obtains diagnostic information from a cluster system file.


ndb_print_sys_file file_name

file_name is the name of a cluster system file (sysfile). Cluster system files are located in a data node's data directory (DataDir); the path under this directory to system files matches the pattern ndb_#_fs/D#/DBDIH/P#.sysfile. In each case, the # represents a number (not necessarily the same number).

Like ndb_print_backup_file and ndb_print_schema_file (and unlike most of the other NDB utilities that are intended to be run on a management server host or to connect to a management server) ndb_print_backup_file must be run on a cluster data node, since it accesses the data node filesystem directly. Because it does not make use of the management server, this utility can be used when the management server is not running, and even when the cluster has been completely shut down.

Additional Options: None.

15.9.10. ndb_select_all

Description: Prints all rows from an NDB table to stdout.


ndb_select_all -c connect_string tbl_name -d db_name [> file_name]

Additional Options:

  • -l lock_type, --lock=lock_type

    Employs a lock when reading the table. Possible values for lock_type are:

    • 0: Read lock

    • 1: Read lock with hold

    • 2: Exclusive read lock

    There is no default value for this option.

  • -o index_name, --order=index_name

    Orders the output according to the index named index_name. Note that this is the name of an index, not of a column, and that the index must have been explicitly named when created.

  • -z, --descending

    Sorts the output in descending order. This option can be used only in conjunction with the -o (--order) option.

  • --header=FALSE

    Excludes column headers from the output.

  • -x, --useHexFormat

    Causes all numeric values to be displayed in hexadecimal format. This does not affect the output of numerals contained in strings or datetime values.

    -D character, --delimiter=character

    Causes the character to be used as a column delimiter. Only table data columns are separated by this delimiter.

    The default delimiter is the tab character.

  • --disk

    Adds a disk reference column to the output. The column is non-empty only for Disk Data tables having non-indexed columns.

  • --rowid

    Adds a ROWID column providing information about the fragments in which rows are stored.

  • --gci

    Adds a column to the output showing the global checkpoint at which each row was last updated. See Section 15.15, “MySQL Cluster Glossary”, and Section, “Log Events”, for more information about checkpoints.

  • -t, --tupscan

    Scan the table in the order of the tuples.

  • --nodata

    Causes any table data to be omitted.

Sample Output:

Output from a MySQL SELECT statement:

mysql> SELECT * FROM;
| id | name      |
|  3 | shark     |
|  6 | puffer    |
|  2 | tuna      |
|  4 | manta ray |
|  5 | grouper   |
|  1 | guppy     |
6 rows in set (0.04 sec)

Output from the equivalent invocation of ndb_select_all:

shell> ./ndb_select_all -c localhost fish -d ctest1
id      name
3       [shark]
6       [puffer]
2       [tuna]
4       [manta ray]
5       [grouper]
1       [guppy]
6 rows returned

NDBT_ProgramExit: 0 - OK

Note that all string values are enclosed by square brackets (“[...]”) in the output of ndb_select_all. For a further example, consider the table created and populated as shown here:

    name VARCHAR(25) NOT NULL,
    breed VARCHAR(50) NOT NULL,
    PRIMARY KEY pk (id),
    KEY ix (name)

    ('', 'Lassie', 'collie'),
    ('', 'Scooby-Doo', 'Great Dane'),
    ('', 'Rin-Tin-Tin', Alsation),
    ('', 'Rosscoe', 'Mutt');

This demonstrates the use of several additional ndb_select_all options:

shell> ./ndb_select_all -d ctest1 dogs -o ix -z --gci --disk        
GCI     id name          breed        DISK_REF
834461  2  [Scooby-Doo]  [Great Dane] [ m_file_no: 0 m_page: 98 m_page_idx: 0 ]
834878  4  [Rosscoe]     [Mutt]       [ m_file_no: 0 m_page: 98 m_page_idx: 16 ]
834463  3  [Rin-Tin-Tin] [Alsation]   [ m_file_no: 0 m_page: 34 m_page_idx: 0 ]
835657  1  [Lassie]      [Collie]     [ m_file_no: 0 m_page: 66 m_page_idx: 0 ]
4 rows returned

NDBT_ProgramExit: 0 - OK

15.9.11. ndb_select_count

Description: Prints the number of rows in one or more NDB table. With a single table, the result is equivalent to that obtained by using the MySQL statement SELECT COUNT(*) FROM tbl_name.


ndb_select_count [-c connect_string] -ddb_name tbl_name[, tbl_name2[, ...]]

Additional Options: None that are specific to this application. However, you can obtain row counts from multiple tables in the same database by listing the table names separated by spaces when invoking this command, as shown under Sample Output.

Sample Output:

shell> ./ndb_select_count -c localhost -d ctest1 fish dogs
6 records in table fish
4 records in table dogs

NDBT_ProgramExit: 0 - OK

15.9.12. ndb_show_tables

Description: Displays a list of all NDB database objects in the cluster. By default, this includes not only both user-created tables and NDB system tables, but NDB-specific indexes, internal triggers, and Cluster Disk Data objects as well.


ndb_show_tables [-c connect_string]

Additional Options:

  • -l, --loops

    Specifies the number of times the utility should execute. This is 1 when this option is not specified, but if you do use the option, you must supply an integer argument for it.

  • -p, --parsable

    Using this option causes the output to be in a format suitable for use with LOAD DATA INFILE.

  • -t, --type

    Can be used to restrict the output to one type of object, specified by an integer type code as shown here:

    • 1: System table

    • 2: User-created table

    • 3: Unique hash index

    Any other value causes all NDB database objects to be listed (the default).

  • -u, --unqualified

    If specified, this causes unqualified object names to be displayed.

Note: Only user-created Cluster tables may be accessed from MySQL; system tables such as SYSTAB_0 are not visible to mysqld. However, you can examine the contents of system tables using NDB API applications such as ndb_select_all (see Section 15.9.10, “ndb_select_all).

15.9.13. — NDBCluster Size Requirement Estimator

This is a Perl script that can be used to estimate the amount of space that would be required by a MySQL database if it were converted to use the NDBCluster storage engine. Unlike the other utilities discussed in this section, it does not require access to a MySQL Cluster (in fact, there is no reason for it to do so). However, it does need to access the MySQL server on which the database to be tested resides.


  • A running MySQL server. The server instance does not have to provide support for MySQL Cluster.

  • A working installation of Perl.

  • The DBI and HTML::Template modules, both of which can be obtained from CPAN if they are not already part of your Perl installation. (Many Linux and other operating system distribution provide their own packages for one or both of these libraries.)

  • The ndb_size.tmpl template file, which you should be able to find in the share/mysql directory of your MySQL installation. This file should be copied or moved into the same directory as — if it is not there already — before running the script.

  • A MySQL user account having the necessary privileges. If you do not wish to use an existing account, then creating one using GRANT USAGE ON db_name.* — where db_name is the name of the database to be examined — is sufficient for this purpose. and ndb_size.tmpl can also be found in the MySQL sources in storage/ndb/tools. If these files are not present in your MySQL installation, you can obtain them from the MySQLForge project page.


perl db_name hostname username password > file_name.html

The command shown connects to the MySQL server at hostname using the account of the user username having the password password, analyses all of the tables in database db_name, and generates a report in HTML format which is directed to the file file_name.html. (Without the redirection, the output is sent to stdout.) This figure shows partial sample output as viewed in a Web browser:

Partial sample output from
     as viewed in a Web

The output from this script includes:

  • Minimum values for the DataMemory, IndexMemory, MaxNoOfTables, MaxNoOfAttributes, MaxNoOfOrderedIndexes, MaxNoOfUniqueHashIndexes, and MaxNoOfTriggers configuration parameters required to accommodate the tables analysed.

  • Memory requirements for all of the tables, attributes, ordered indexes, and unique hash indexes defined in the database.

  • The IndexMemory and DataMemory required per table and table row.

15.9.14. ndb_waiter

Description: Repeatedly (each 100 milliseconds) prints out the status of all cluster data nodes until either the cluster reaches a given status or the --timeout limit is exceeded, then exits. By default, it waits for the cluster to achieve STARTED status, in which all nodes have started and connected to the cluster. This can be overridden using the --no-contact and --not-started options (see Additional Options).

The node states reported by this utility are as follows:

  • NO_CONTACT: The node cannot be contacted.

  • UNKNOWN: The node can be contacted, but its status is not yet known. Usually, this means that the node has received a START or RESTART command from the management server, but has not yet acted on it.

  • NOT_STARTED: The node has stopped, but remains in contact with the cluster. This is seen when restarting the node using the management client's RESTART command.

  • STARTING: The node's ndbd process has started, but the node has not yet joined the cluster.

  • STARTED: The node is operational, and has joined the cluster.

  • SHUTTING_DOWN: The node is shutting down.

  • SINGLE USER MODE: This is shown for all cluster data nodes when the cluster is in single user mode.


ndb_waiter [-c connect_string]

Additional Options:

  • -n, --no-contact

    Instead of waiting for the STARTED state, ndb_waiter continues running until the cluster reaches NO_CONTACT status before exiting.

  • --not-started

    Instead of waiting for the STARTED state, ndb_waiter continues running until the cluster reaches NOT_STARTED status before exiting.

  • -t, --timeout=#

    Time to wait. The program exits if the desired state is not achieved within this number of seconds. The default is 120 seconds (1200 reporting cycles).

Sample Output:

Shown here is the output from ndb_waiter when run against a 4-node cluster in which two nodes have been shut down and then started again manually. Duplicate reports (indicated by “...”) are omitted.

shell> ./ndb_waiter -c localhost

Connecting to mgmsrv at (localhost)
State node 1 STARTED
State node 2 NO_CONTACT
State node 3 STARTED
State node 4 NO_CONTACT
Waiting for cluster enter state STARTED


State node 1 STARTED
State node 2 UNKNOWN
State node 3 STARTED
State node 4 NO_CONTACT
Waiting for cluster enter state STARTED


State node 1 STARTED
State node 2 STARTING
State node 3 STARTED
State node 4 NO_CONTACT
Waiting for cluster enter state STARTED


State node 1 STARTED
State node 2 STARTING
State node 3 STARTED
State node 4 UNKNOWN
Waiting for cluster enter state STARTED


State node 1 STARTED
State node 2 STARTING
State node 3 STARTED
State node 4 STARTING
Waiting for cluster enter state STARTED


State node 1 STARTED
State node 2 STARTED
State node 3 STARTED
State node 4 STARTING
Waiting for cluster enter state STARTED


State node 1 STARTED
State node 2 STARTED
State node 3 STARTED
State node 4 STARTED
Waiting for cluster enter state STARTED

NDBT_ProgramExit: 0 - OK

Note: If no connectstring is specified, then ndb_waiter tries to connect to a management on localhost, and reports Connecting to mgmsrv at (null).

15.10. MySQL Cluster Replication

Previous to MySQL 5.1.6, asynchronous replication, more usually referred to simply as “replication”, was not available when using MySQL Cluster. MySQL 5.1.6 introduces master-slave replication of this type for MySQL Cluster databases. This section explains how to set up and manage a configuration wherein one group of computers operating as a MySQL Cluster replicates to a second computer or group of computers. We assume some familiarity on the part of the reader with standard MySQL replication as discussed elsewhere in this Manual. (See Chapter 6, Replication).

Normal (non-clustered) replication involves a “master” server and a “slave” server, the master being the source of the operations and data to be replicated and the slave being the recipient of these. In MySQL Cluster, replication is conceptually very similar but can be more complex in practice, as it may be extended to cover a number of different configurations including replicating between two complete clusters. Although a MySQL Cluster itself depends on the NDB Cluster storage engine for clustering functionality, it is not necessary to use the Cluster storage engine on the slave. However, for maximum availability, it is possible to replicate from one MySQL Cluster to another, and it is this type of configuration that we discuss, as shown in the following figure:

MySQL Cluster-to-Cluster Replication

In this scenario, the replication process is one in which successive states of a master cluster are logged and saved to a slave cluster. This process is accomplished by a special thread known as the NDB binlog injector thread, which runs on each MySQL server and produces a binary log (binlog). This thread ensures that all changes in the cluster producing the binary log — and not just those changes that are effected via the MySQL Server — are inserted into the binary log with the correct serialization order. We refer to the MySQL replication master and replication slave servers as replication servers or replication nodes, and the data flow or line of communication between them as a replication channel.

15.10.1. Abbreviations and Symbols

Throughout this section, we use the following abbreviations or symbols for referring to the master and slave clusters, and to processes and commands run on the clusters or cluster nodes:

Symbol or AbbreviationDescription (Refers to...)
MThe cluster serving as the (primary) replication master
SThe cluster acting as the (primary) replication slave
shellM>Shell command to be issued on the master cluster
mysqlM>MySQL client command issued on a single MySQL server running as an SQL node on the master cluster
mysqlM*>MySQL client command to be issued on all SQL nodes participating in the replication master cluster
shellS>Shell command to be issued on the slave cluster
mysqlS>MySQL client command issued on a single MySQL server running as an SQL node on the slave cluster
mysqlS*>MySQL client command to be issued on all SQL nodes participating in the replication slave cluster
CPrimary replication channel
C'Secondary replication channel
M'Secondary replication master
S'Secondary replication slave

15.10.2. Assumptions and General Requirements

A replication channel requires two MySQL servers acting as replication servers (one each for the master and slave). For example, this means that in the case of a replication setup with two replication channels (to provide an extra channel for redundancy), there will be a total of four replication nodes, two per cluster.

Replication of a MySQL Cluster as described in this section and those following is dependent on row-based replication. This means that the replication master MySQL server must be started with --binlog-format=row, as described in Section 15.10.6, “Starting Replication (Single Replication Channel)”. For general information about row-based replication, see Section 6.1.2, “Replication Formats”.

(It is possible to replicate a MySQL Cluster using statement-based replication. However, in this case, the following restrictions apply: All updates to data rows on the cluster acting as the master must be directed to a single MySQL server; It is not possible to replicate a cluster using several MySQL replication processes at the same time; Only changes made at the SQL level are replicated.)

Each MySQL server used for replication in either cluster must be uniquely identified among all the MySQL replication servers participating in either cluster (you cannot have replication servers on both the master and slave clusters sharing the same ID). This can be done by starting each SQL node using the --server-id=id option, where id is a unique integer. Although it is not strictly necessary, we will assume for purposes of this discussion that all MySQL installations are the same version.

In any event, both MySQL servers involved in replication must be compatible with one another with respect to both the version of the replication protocol used and the SQL feature sets which they support; the simplest and easiest way to assure that this is the case is to use the same MySQL version for all servers involved. Note that in many cases it is not possible to replicate to a slave running a version of MySQL with a lower version number than that of the master — see Section 6.3.2, “Replication Compatibility Between MySQL Versions”, for details.

We assume that the slave server or cluster is dedicated to replication of the master, and that no other data is being stored on it.

15.10.3. Known Issues

The following are known problems or issues when using replication with MySQL Cluster in MySQL 5.1:

  • A MySQL Cluster replication slave mysqld has no way of detecting that the connection from the master has been interrupted, and no buffering of the log takes place. For this reason, if a problem such as the master going down or a network failure occurring, then it is possible for the slave to become inconsistent with the master.

    To guard against this possibility, it is advisable to set up multiple replication lines, to monitor the master mysqld on the primary replication line, and to fail over to a secondary line if and as necessary.

    For information about implementing this type of setup, see Section 15.10.7, “Using Two Replication Channels”, and Section 15.10.8, “Implementing Failover with MySQL Cluster”.

    We are working on an integrated solution to this problem for inclusion in a future MySQL Cluster release. (See the discussion of this issue in Bug#21494.)

  • Circular replication is not supported for Cluster replication. This is because all log events created in a particular MySQL Cluster are wrongly tagged with the server id of the MySQL server used as master and not with the server id of the originating server.

    Because of this misidentification, replicating between MySQL servers A→B→A, where B is any MySQL server connected to a cluster A, causes the cluster table changes (log entries) of A to “lose” the originating server's identifier in the log from B to A. This causes the changes to be reapplied on server A.

  • The use of data definition statements, such as CREATE TABLE, DROP TABLE, and ALTER TABLE, are recorded in the binary log for only the MySQL server on which they are issued.

  • In MySQL 5.1.6, only those NDB tables having explicit primary keys could be replicated. This limitation was lifted in MySQL 5.1.7.

  • Restarting the cluster with the --initial option will cause the sequence of GCI and epoch numbers to start over from 0. (This is generally true of MySQL Cluster and not limited to replication scenarios involving Cluster.) The MySQL servers involved in replication should in this case be replicated. After this, you should use the RESET MASTER and RESET SLAVE statements to clear the invalid ndb_binlog_index and ndb_apply_status tables. respectively.

  • Trying to set values for the auto_increment_offset and auto_increment_increment server system variables produces unpredictable results. The use of these variables is not supported with MySQL Cluster replication.

15.10.4. Replication Schema and Tables

Replication in MySQL Cluster makes use of a number of dedicated tables in the mysql database on each MySQL Server instance acting as an SQL node in both the cluster being replicated and the replication slave (whether the slave is a single server or a cluster). These tables are created during the MySQL installation process by the mysql_install_db script, and include a table for storing the binary log's indexing data. Since the ndb_binlog_index table is local to each MySQL server and does not participate in clustering, it uses the MyISAM storage engine. This means that it must be created separately on each mysqld participating in the master cluster. (However, the binlog itself contains updates from all MySQL servers in the cluster to be replicated.) This table is defined as follows:

CREATE TABLE `ndb_binlog_index` (
          `Position`  BIGINT(20) UNSIGNED NOT NULL,
          `File`      VARCHAR(255) NOT NULL,
          `epoch`     BIGINT(20) UNSIGNED NOT NULL,
          `inserts`   BIGINT(20) UNSIGNED NOT NULL,
          `updates`   BIGINT(20) UNSIGNED NOT NULL,
          `deletes`   BIGINT(20) UNSIGNED NOT NULL,
          `schemaops` BIGNINT(20) UNSIGNED NOT NULL,
          PRIMARY KEY (`epoch`)

Important: Prior to MySQL 5.1.14, the ndb_binlog_index table was known as binlog_index, and was kept in a separate cluster database, which in MySQL 5.1.7 and earlier was known as the cluster_replication database. Similarly, the ndb_apply_status and ndb_schema tables were known as apply_status and schema, and were also found in the cluster (earlier cluster_replication) database. However, beginning with MySQL 5.1.14, all MySQL Cluster replication tables reside in the mysql system database.

Information about how this change affects upgrades from MySQL Cluster 5.1.13 and earlier to 5.1.14 and later versions can be found in Section E.1.3, “Changes in release 5.1.14 (05 December 2006)”.

The following figure shows the relationship of the MySQL Cluster replication master server, its binlog injector thread, and the mysql.ndb_binlog_index table.

The replication master cluster, the
          binlog-injector thread, and the
          ndb_binlog_index table

An additional table, named ndb_apply_status, is used to keep a record of the operations that have been replicated from the master to the slave. Unlike the case with ndb_binlog_index, the data in this table is not specific to any one SQL node in the (slave) cluster, and so ndb_apply_status can use the NDB Cluster storage engine, as shown here:

CREATE TABLE `ndb_apply_status` (
    `server_id` INT(10) UNSIGNED NOT NULL,
    `epoch`     BIGINT(20) UNSIGNED NOT NULL,
    PRIMARY KEY  USING HASH (`server_id`)

The ndb_binlog_index and ndb_apply_status tables are created in the mysql database because they should not be replicated. No user intervention is normally required to create or maintain either of them. Both the ndb_binlog_index and the ndb_apply_status tables are maintained by the NDB injector thread. This keeps the master mysqld process updated to changes performed by the NDB storage engine. The NDB binlog injector thread receives events directly from the NDB storage engine. The NDB injector is responsible for capturing all the data events within the cluster, and ensures that all events which change, insert, or delete data are recorded in the ndb_binlog_index table. The slave I/O thread transfers the events from the master's binary log to the slave's relay log.

However, it is advisable to check for the existence and integrity of these tables as an initial step in preparing a MySQL Cluster for replication. It is possible to view event data recorded in the binary log by querying the mysql.binlog_index table directly on the master. This can be also be accomplished using the SHOW BINLOG EVENTS statement on either the replication master or slave MySQL servers.

Note: Beginning with MySQL 5.1.14, if the apply_status table does not exist on the slave, it is created by ndb_restore. (Bug#14612)

15.10.5. Preparing the Cluster for Replication

Preparing the MySQL Cluster for replication consists of the following steps:

  1. Check all MySQL servers for version compatibility (see Section 15.10.2, “Assumptions and General Requirements”).

  2. Create a slave account on the master Cluster with the appropriate privileges:

         -> ON *.* TO 'slave_user'@'slave_host'
         -> IDENTIFIED BY 'slave_password';

    where slave_user is the slave account username, slave_host is the hostname or IP address of the replication slave, and slave_password is the password to assign to this account.

    For example, to create a slave user account with the name “myslave,” logging in from the host named “rep-slave,” and using the password “53cr37,” use the following GRANT statement:

         -> ON *.* TO 'myslave'@'rep-slave'
         -> IDENTIFIED BY '53cr37';

    For security reasons, it is preferable to use a unique user account — not employed for any other purpose — for the replication slave account.

  3. Configure the slave to use the master. Using the MySQL Monitor, this can be accomplished with the CHANGE MASTER TO statement:

         -> MASTER_HOST='master_host',
         -> MASTER_PORT=master_port,
         -> MASTER_USER='slave_user',
         -> MASTER_PASSWORD='slave_password';

    where master_host is the hostname or IP address of the replication master, master_port is the port for the slave to use for connecting to the master, slave_user is the username set up for the slave on the master, and slave_password is the password set for that user account in the previous step.

    For example, to tell the slave to replicate from the MySQL server whose hostname is “rep-master,” using the replication slave account created in the previous step, use the following statement:

         -> MASTER_HOST='rep-master'
         -> MASTER_PORT=3306,
         -> MASTER_USER='myslave'
         -> MASTER_PASSWORD='53cr37';

    (For a complete list of clauses that can be used with this statement, see Section, “CHANGE MASTER TO Syntax”.)

    You can also configure the slave to use the master by setting the corresponding startup options in the slave server's my.cnf file. To configure the slave in the same way as the preceding example CHANGE MASTER TO statement, the following information would need to be included in the slave's my.cnf file:


    See Section 6.1.3, “Replication Options and Variables”, for additional options that can be set in my.cnf for replication slaves.

    Note: To provide replication backup capability, you will also need to add an ndb-connectstring option to the slave's my.cnf file prior to starting the replication process. See Section 15.10.9, “MySQL Cluster Backups With Replication”, for details.

  4. If the master cluster is already in use, you can create a backup of the master and load this onto the slave to cut down on the amount of time required for the slave to synchronize itself with the master. If the slave is also running MySQL Cluster, this can be accomplished using the backup and restore procedure described in Section 15.10.9, “MySQL Cluster Backups With Replication”.


    In the event that you are not using MySQL Cluster on the replication slave, you can create a backup with this command on the replication master:

    shellM> mysqldump --master-data=1

    Then import the resulting data dump onto the slave by copying the dump file over to the slave. After this, you can use the mysql client to import the data from the dumpfile into the slave database as shown here, where dump_file is the name of the file that was generated using mysqldump on the master, and db_name is the name of the database to be replicated:

    shellS> mysql -u root -p db_name < dump_file

    For a complete list of options to use with mysqldump, see Section 8.12, “mysqldump — A Database Backup Program”.

    Note that if you copy the data to the slave in this fashion, you should make sure that the slave is started with the --skip-slave-start option on the command line, or else include skip-slave-start in the slave's my.cnf file to keep it from trying to connect to the master to begin replicating before all the data has been loaded. Once the data loading has completed, follow the additional steps outlined in the next two sections.

  5. Ensure that each MySQL server acting as a replication master is configured with a unique server ID, and with binary logging enabled, using the row format. (See Section 6.1.2, “Replication Formats”.) These options can be set either in the master server's my.cnf file, or on the command line when starting the master mysqld process. See Section 15.10.6, “Starting Replication (Single Replication Channel)”, for information regarding the latter option.

15.10.6. Starting Replication (Single Replication Channel)

This section outlines the procedure for starting MySQL CLuster replication using a single replication channel.

  1. Start the MySQL replication master server by issuing this command:

    shellM> mysqld --nbdcluster --server-id=id \ 
            --log-bin --binlog-format=row &

    where id is this server's unique ID (see Section 15.10.2, “Assumptions and General Requirements”). This starts the server's mysqld process with binary logging enabled using the proper logging format.

  2. Start the MySQL replication slave server as shown here:

    shellS> mysqld --ndbcluster --server-id=id &

    where id is the slave server's unique ID. It is not necessary to enable logging on the replication slave.

    Note that you should use the --skip-slave-start option with this command or else you should include skip-slave-start in the slave server's my.cnf file, unless you want replication to begin immediately. With the use of this option, the start of replication is delayed until the appropriate START SLAVE statement has been issued, as explained in Step 4 below.

  3. It is necessary to synchronize the slave server with the master server's replication binlog. If binary logging has not previously been running on the master, run the following statement on the slave:

         -> MASTER_LOG_FILE='',
         -> MASTER_LOG_POS=4;

    This instructs the slave to begin reading the master's binary log from the log's starting point. Otherwise — that is, if you are loading data from the master using a backup — see Section 15.10.8, “Implementing Failover with MySQL Cluster”, for information on how to obtain the correct values to use for MASTER_LOG_FILE and MASTER_LOG_POS in such cases.

  4. Finally, you must instruct the slave to begin applying replication by issuing this command from the mysql client on the replication slave:

    mysqlS> START SLAVE;

    This also initiates the transmission of replication data from the master to the slave.

It is also possible to use two replication channels, in a manner simlar to the procedure described in the next section; the differences between this and using a single replication channel are covered in Section 15.10.7, “Using Two Replication Channels”.

15.10.7. Using Two Replication Channels

In a more complete example scenario, we envision two replication channels to provide redundancy and thereby guard against possible failure of a single replication channel. This requires a total of four replication servers, two masters for the master cluster and two slave servers for the slave cluster. For purposes of the discussion that follows, we assume that unique identifiers are assigned as shown here:

Server IDDescription
1Master - primary replication channel (M)
2Master - secondary replication channel (M')
3Slave - primary replication channel (S)
4Slave - secondary replication channel (S')

Setting up replication with two channels is not radically different from setting up a single replication channel. First, the mysqld processes for the primary and secondary replication masters must be started, followed by those for the primary and secondary slaves. Then the replication processes may be initiated by issuing the START SLAVE statement on each of the slaves. The commands and the order in which they need to be issued are shown here:

  1. Start the primary replication master:

    shellM> mysqld --ndbcluster --server-id=1 \ 
                   --log-bin --binlog-format=row &
  2. Start the secondary replication master:

    shellM'> mysqld --ndbcluster --server-id=2 \
                   --log-bin --binlog-format=row &
  3. Start the primary replication slave server:

    shellS> mysqld --ndbcluster --server-id=3 \
                   --skip-slave-start &
  4. Start the secondary replication slave:

    shellS'> mysqld --ndbcluster --server-id=4 \
                    --skip-slave-start &
  5. Finally, initiate replication on the primary channel by executing the START SLAVE statement on the primary slave as shown here:

    mysqlS> START SLAVE;

As mentioned previously, it is not necessary to enable binary logging on replication slaves.

15.10.8. Implementing Failover with MySQL Cluster

In the event that the primary Cluster replication process fails, it is possible to switch over to the secondary replication channel. The following procedure describes the steps required to accomplish this.

  1. Obtain the time of the most recent global checkpoint (GCP). That is, you need to determine the most recent epoch from the ndb_apply_status table on the slave cluster, which can be found using the following query:

    mysqlS'> SELECT @latest:=MAX(epoch)
          ->        FROM mysql.ndb_apply_status;
  2. Using the information obtained from the query shown in Step 1, obtain the corresponding records from the ndb_binlog_index table on the master cluster as shown here:

    mysqlM'> SELECT 
          ->     @file:=SUBSTRING_INDEX(File, '/', -1),
          ->     @pos:=Position
          -> FROM mysql.ndb_binlog_index
          -> WHERE epoch > @latest
          -> ORDER BY epoch ASC LIMIT 1;

    These are the records saved on the master since the failure of the primary replication channel. We have employed a user variable @latest here to represent the value obtained in Step 1. Of course, it is not possible for one mysqld instance to access user variables set on another server instance directly. These values must be “plugged in” to the second query manually or in application code.

  3. Now it is possible to synchronize the secondary channel by running the following query on the secondary slave server:

          ->     MASTER_LOG_FILE='@file',
          ->     [email protected];

    Again we have employed user variables (in this case @file and @pos) to represent the values obtained in Step 2 and applied in Step 3; in practice these values must be inserted manually or using application code that can access both of the servers involved.

    Note that @file is a string value such as '/var/log/mysql/replication-master-bin.00001' and so must be quoted when used in SQL or application code. However, the value represented by @pos must not be quoted. Although MySQL normally attempts to convert strings to numbers, this case is an exception.

  4. You can now initiate replication on the secondary channel by issuing the appropriate command on the secondary slave mysqld:

    mysqlS'> START SLAVE;

Once the secondary replication channel is active, you can investigate the failure of the primary and effect repairs. The precise actions required to do this will depend upon the reasons for which the primary channel failed.

If the failure is limited to a single server, it should (in theory) be possible to replicate from M to S', or from M' to S; however, this has not yet been tested.

15.10.9. MySQL Cluster Backups With Replication

This section discusses making backups and restoring from them using MySQL Cluster replication. We assume that the replication servers have already been configured as covered previously (see Section 15.10.5, “Preparing the Cluster for Replication”, and the sections immediately following). This having been done, the procedure for making a backup and then restoring from it is as follows:

  1. There are two different methods by which the backup may be started.

    • Method A:

      This method requires that the cluster backup process was previously enabled on the master server, prior to starting the replication process. This can be done by including the line


      in a [MYSQL_CLUSTER] section in the my.cnf file, where management_host is the IP address or hostname of the NDB management server for the master cluster, and port is the management server's port number. Note that the port number needs to be specified only if the default port (1186) is not being used. (See Section 15.3.3, “Multi-Computer Configuration”, for more information about ports and port allocation in MySQL Cluster.)

      In this case, the backup can be started by executing this statement on the replication master:

      shellM> ndb_mgm -e "START BACKUP"
    • Method B:

      If the my.cnf file does not specify where to find the management host, you can start the backup process by passing this information to the NDB management client as part of the START BACKUP command, like this:

      shellM> ndb_mgm management_host:port -e "START BACKUP"

      where management_host and port are the hostname and port number of the management server. In our scenario as outlined earlier (see Section 15.10.5, “Preparing the Cluster for Replication”), this would be executed as follows:

      shellM> ndb_mgm rep-master:1186 -e "START BACKUP"

    In either case, it is highly advisable to allow any pending transactions to be completed before beginning the backup, and then not to permit any new transactions to begin during the backup process.

  2. Copy the cluster backup files to the slave that is being brought on line. Each system running an ndbd process for the master cluster will have cluster backup files located on it, and all of these files must be copied to the slave to ensure a successful restore. The backup files can be copied into any directory on the computer where the slave management host resides, so long as the MySQL and NDB binaries have read permissions in that directory. In this case, we will assume that these files have been copied into the directory /var/BACKUPS/BACKUP-1.

    It is not necessary that the slave cluster have the same number of ndbd processes (data nodes) as the master; however, it is highly recommended this number be the same. It is necessary that the slave be started with the --skip-slave-start option, to prevent premature startup of the replication process.

  3. Create any databases on the slave cluster that are present on the master cluster that are to be replicated to the slave. Important: A CREATE SCHEMA statement corresponding to each database to be replicated must be executed on each data node in the slave cluster.

  4. Reset the slave cluster using this statement in the MySQL Monitor:

    mysqlS> RESET SLAVE;

    It is important to make sure that the slave's apply_status table does not contain any records prior to running the restore process. You can accomplish this by running this SQL statement on the slave:

    mysqlS> DELETE FROM mysql.ndb_apply_status;
  5. You can now start the cluster restoration process on the replication slave using the ndb_restore command for each backup file in turn. For the first of these, it is necessary to include the -m option to restore the cluster metadata:

    shellS> ndb_restore -c slave_host:port -n node-id \
            -b backup-id -m -r dir

    dir is the path to the directory where the backup files have been placed on the replication slave. For the ndb_restore commands corresponding to the remaining backup files, the -m option should not be used.

    For restoring from a master cluster with four data nodes (as shown in the figure in Section 15.10, “MySQL Cluster Replication”) where the backup files have been copied to the directory /var/BACKUPS/BACKUP-1, the proper sequence of commands to be executed on the slave might look like this:

    shellS> ndb_restore -c rep-slave:1186 -n 2 -b 1 -m \
            -r ./VAR/BACKUPS/BACKUP-1
    shellS> ndb_restore -c rep-slave:1186 -n 3 -b 1 \
            -r ./VAR/BACKUPS/BACKUP-1
    shellS> ndb_restore -c rep-slave:1186 -n 4 -b 1 \
            -r ./VAR/BACKUPS/BACKUP-1
    shellS> ndb_restore -c rep-slave:1186 -n 5 -b 1 -e \
            -r ./VAR/BACKUPS/BACKUP-1

    This sequence of commands causes the most recent epoch records to be written to the slave's ndb_apply_status table.

  6. Now you need to obtain the most recent epoch from the ndb_binlog_index table on the slave (as discussed in Section 15.10.8, “Implementing Failover with MySQL Cluster”):

    mysqlS> SELECT @latest:=MAX(epoch)
            FROM mysql.ndb_binlog_index;
  7. Using @latest as the epoch value obtained in the previous step, you can obtain the correct starting position @pos in the correct binary log file @file from the master's mysql.ndb_binlog_index table using the query shown here:

    mysqlM> SELECT 
         ->     @file:=SUBSTRING_INDEX(File, '/', -1),
         ->     @pos:=Position
         -> FROM mysql.ndb_binlog_index
         -> WHERE epoch > @latest
         -> ORDER BY epoch ASC LIMIT 1;
  8. Using the values obtained in the previous step, you can now issue the appropriate CHANGE MASTER TO statement in the slave's mysql client:

         ->     MASTER_LOG_FILE='@file',
         ->     [email protected];
  9. Now that the slave “knows” from what point in which binlog file to start reading data from the master, you can cause the slave to begin replicating with this standard MySQL statement:

    mysqlS> START SLAVE;

To perform a backup and restore on a second replication channel, it is necessary only to repeat these steps, substituting the hostnames and IDs of the secondary master and slave for those of the primary master and slave replication servers where appropriate, and running the preceding statements on them.

For additional information on performing Cluster backups and restoring Cluster from backups, see Section 15.8, “On-line Backup of MySQL Cluster”. Automating Synchronization of the Slave to the Master binlog

It is possible to automate much of the process described in the previous section (see Section 15.10.9, “MySQL Cluster Backups With Replication”). The following Perl script serves as an example of how you can do this.

#!/user/bin/perl -w

#  file:

#  Copyright ©2005 MySQL AB

#  This program is free software; you can redistribute it and/or modify
#  it under the terms of the GNU General Public License as published by
#  the Free Software Foundation; either version 2 of the License, or
#  (at your option) any later version.

#  This program is distributed in the hope that it will be useful,
#  but WITHOUT ANY WARRANTY; without even the implied warranty of
#  GNU General Public License for more details.

#  You should have received a copy of the GNU General Public License
#  along with this program; if not, write to: 
#  Free Software Foundation, Inc. 
#  59 Temple Place, Suite 330 
#  Boston, MA 02111-1307 USA
#  Version 1.1

######################## Includes ###############################

use DBI;

######################## Globals ################################

my  $m_host='';
my  $m_port='';
my  $m_user='';
my  $m_pass='';
my  $s_host='';
my  $s_port='';
my  $s_user='';
my  $s_pass='';
my  $dbhM='';
my  $dbhS='';

####################### Sub Prototypes ##########################

sub CollectCommandPromptInfo;
sub ConnectToDatabases;
sub DisconnectFromDatabases;
sub GetSlaveEpoch;
sub GetMasterInfo;
sub UpdateSlave;

######################## Program Main ###########################


################## Collect Command Prompt Info ##################

sub CollectCommandPromptInfo
  ### Check that user has supplied correct number of command line args
  die "Usage:\n
       reset-slave >master MySQL host< >master MySQL port< \n
                   >master user< >master pass< >slave MySQL host< \n
                   >slave MySQL port< >slave user< >slave pass< \n
       All 8 arguments must be passed. Use BLANK for NULL passwords\n"
       unless @ARGV == 8;

  $m_host  =  $ARGV[0];
  $m_port  =  $ARGV[1];
  $m_user  =  $ARGV[2];
  $m_pass  =  $ARGV[3];
  $s_host  =  $ARGV[4];
  $s_port  =  $ARGV[5];
  $s_user  =  $ARGV[6];
  $s_pass  =  $ARGV[7];

  if ($m_pass eq "BLANK") { $m_pass = '';}
  if ($s_pass eq "BLANK") { $s_pass = '';}

###############  Make connections to both databases #############

sub ConnectToDatabases
  ### Connect to both master and slave cluster databases

  ### Connect to master
    = DBI->connect(
    "$m_user", "$m_pass")
      or die "Can't connect to Master Cluster MySQL process!
              Error: $DBI::errstr\n";

  ### Connect to slave
    = DBI->connect(
          "$s_user", "$s_pass")
    or die "Can't connect to Slave Cluster MySQL process!
            Error: $DBI::errstr\n";

################  Disconnect from both databases ################

sub DisconnectFromDatabases
  ### Disconnect from master

  or warn " Disconnection failed: $DBI::errstr\n";

  ### Disconnect from slave

  or warn " Disconnection failed: $DBI::errstr\n";

######################  Find the last good GCI ##################

sub GetSlaveEpoch
  $sth = $dbhS->prepare("SELECT MAX(epoch)
                         FROM mysql.ndb_apply_status;")
      or die "Error while preparing to select epoch from slave: ", 

      or die "Selecting epoch from slave error: ", $sth->errstr;

  $sth->bind_col (1, \$epoch);
  print "\tSlave Epoch =  $epoch\n";

#######  Find the position of the last GCI in the binlog ########

sub GetMasterInfo
  $sth = $dbhM->prepare("SELECT 
                           SUBSTRING_INDEX(File, '/', -1), Position
                         FROM mysql.ndb_binlog_index
                         WHERE epoch > $epoch 
                         ORDER BY epoch ASC LIMIT 1;")
      or die "Prepare to select from master error: ", $dbhM->errstr;

      or die "Selecting from master error: ", $sth->errstr;

  $sth->bind_col (1, \$binlog);
  $sth->bind_col (2, \$binpos);
  print "\tMaster bin log =  $binlog\n";
  print "\tMaster Bin Log position =  $binpos\n";

##########  Set the slave to process from that location #########

sub UpdateSlave
  $sth = $dbhS->prepare("CHANGE MASTER TO
      or die "Prepare to CHANGE MASTER error: ", $dbhS->errstr;

       or die "CHNAGE MASTER on slave error: ", $sth->errstr;
  print "\tSlave has been updated. You may now start the slave.\n";

# end

15.11. MySQL Cluster Disk Data Storage

In MySQL 5.1.6, it became possible to store the non-indexed columns of NDB tables on disk, rather than in RAM as with previous versions of MySQL Cluster.

As part of implementing Cluster Disk Data work, a number of improvements were made in MySQL Cluster for the efficient handling of very large amounts (terabytes) of data during node recovery and restart. These include a “no-steal” algorithm for synchronising a starting node with very large data sets. For more information, see the paper Recovery Principles of MySQL Cluster 5.1, by MySQL Cluster developers Mikael Ronström and Jonas Oreland.

Assuming that you have already set up a MySQL Cluster with all nodes (including management and SQL nodes) running MySQL 5.1.6 or newer, the basic steps for creating a Cluster table on disk are as follows:

  1. Create a log file group, and assign one or more UNDO log files to it (an UNDO log file is also referred as an UNDOFILE).

  2. Create a tablespace, and assign the log file group to it, as well as one or more data files.

  3. Create a Disk Data table that uses this tablespace for data storage.

Each of these tasks can be accomplished using SQL statements, as shown in the following example.

  1. We create a log file group named lg_1 using CREATE LOGFILE GROUP. This log file group is to be made up of two UNDO log files, which we name undo_1.dat and undo_2.dat, whose initial sizes are 16 MB and 12 MB, respectively. (You must specify a log file's initial size when adding it to a log file group.) Optionally, you can also specify a size for the log file group's UNDO buffer, or allow it to assume the default value of 8 MB. In this example, we set the UNDO buffer's size at 2 MB. A log file group must be created with an UNDO log file; so we add undo_1.dat to lg_1 in this CREATE LOGFILE GROUP statement:

        ADD UNDOFILE 'undo_1.dat'
        INITIAL_SIZE 16M
        ENGINE NDB;

    To add undo_2.dat to the log file group, use the following ALTER LOGFILE GROUP statement:

        ADD UNDOFILE 'undo_2.dat'
        INITIAL_SIZE 12M
        ENGINE NDB;

    Some items of note:

    • The .dat file extension used here is not required. We use it merely to make the log and data files easily recognisable.

    • Every CREATE LOGFILE GROUP and ALTER LOGFILE GROUP statement must include an ENGINE clause. In MySQL 5.1, the permitted values for this clause are NDB and NDBCLUSTER.

      Important: In MySQL 5.1.8 and later, there can exist only one log file group at any given time.

    • When you add an UNDO log file to a log file group using ADD UNDOFILE 'filename', a file with the the name filename is created in the ndb_nodeid_fs directory within the DataDirectory of each data node in the cluster, where nodeid is the node ID of the data node.

    • UNDO_BUFFER_SIZE is limited by the amount of system memory available.

    • For more information about the CREATE LOGFILE GROUP statement, see Section 13.1.9, “CREATE LOGFILE GROUP Syntax”. For more information about ALTER LOGFILE GROUP, see Section 13.1.3, “ALTER LOGFILE GROUP Syntax”.

  2. Now we can create a tablespace. A tablespace contains files to be used by MySQL Cluster Disk Data tables for storing their data. A tablespace is also associated with a particular log file group. When creating a new tablespace, you must specify the log file group which it is to use for UNDO logging; you must also specify a data file. You can add more data files to the tablespace after it the tablespace is created; it is also possible to drop data files from a tablespace (an example of dropping data files is provided later in this section).

    Assume that we wish to create a tablespace named ts_1 which uses lg_1 as its log file group. This tablespace is to contain two data files named data_1.dat and data_2.dat, whose initial sizes are 32 MB and 48 MB, respectively. We can do this using two SQL statements: CREATE TABLESPACE, to create ts_1 with the data file data_1.dat, and to associate ts_1 with log file group lg_1; and ALTER TABLESPACE, to add the second data file. We show these statements here:

        ADD DATAFILE 'data_1.dat'
        USE LOGFILE GROUP lg_1
        INITIAL_SIZE 32M
        ENGINE NDB;
        ADD DATAFILE 'data_2.dat'
        INITIAL_SIZE 48M
        ENGINE NDB;

    Some items of note:

    • As is the case with the filenames used here for UNDO log files, there is no special significance for the .dat file extension; it is used merely for easy recognition.

    • All CREATE TABLESPACE and ALTER TABLESPACE statements must contain an ENGINE clause; only tables using the same storage engine as the tablespace can be created in the tablespace. In MySQL 5.1, the only permitted values for this clause are NDB and NDBCLUSTER.

      For more information about the CREATE TABLESPACE and ALTER TABLESPACE statements, see Section 13.1.10, “CREATE TABLESPACE Syntax”, and Section 13.1.4, “ALTER TABLESPACE Syntax”.

  3. Now it is possible to create a table whose non-indexed columns are stored on disk in the tablespace ts_1:

    CREATE TABLE dt_1 (
        last_name VARCHAR(50) NOT NULL,
        first_name VARCHAR(50) NOT NULL,
        dob DATE NOT NULL,
        joined DATE NOT NULL,
        INDEX(last_name, first_name)
        ENGINE NDB;

    The TABLESPACE ... STORAGE DISK clause tells the NDB Cluster storage engine to use tablespace ts_1 for disk data storage.

    Once table ts_1 has been created as shown, you can perform INSERT, SELECT, UPDATE, and DELETE statements on it just as you would with any other MySQL table.

    For table dt_1 as it has been defined here, only the dob and joined columns are stored on disk. This is because there are indexes on the id, last_name, and first_name columns, and so data belonging to these columns is stored in RAM. In MySQL 5.1, only non-indexed columns can be held on disk; indexes and indexed column data continue to be stored in memory. This trade-off between the use of indexes and conservation of RAM is something you must keep in mind as you design Disk Data tables.

    Important: For Disk Data tables in MySQL 5.1, variable-length columns take up a fixed amount of space. For each row, this is equal to the space required to store the largest possible value for that column. (For help in calculating these values, see Section 11.5, “Data Type Storage Requirements”.)

    Note: Starting the cluster with the --initial option does not remove Disk Data files. You must do so manually prior to performing an initial restart of the cluster.

A log file group, a tablespace, and any Disk Data tables using these must be created in a particular order. The same is true for dropping any of these objects:

  • A log file group cannot be dropped, so long as any tablespaces are using it.

  • A tablespace cannot be dropped, so long as it contains any data files.

  • You cannot drop any data files from a tablespace, so long as there remain any tables which are using the tablespace.

  • Beginning with MySQL 5.1.12, it is no longer possible to drop files created in association with a different tablespace than the one with which the files were created. (Bug#20053)

For example, to drop all the objects created so far in this section, you would use the following statements:

mysql> DROP TABLE dt_1;

    -> DROP DATAFILE 'data_2.dat'
    -> ENGINE NDB;

    -> DROP DATAFILE 'data_1.dat'
    -> ENGINE NDB;

    -> ENGINE NDB;

    -> ENGINE NDB;

These statements must be performed in the order shown, except that the two ALTER TABLESPACE ... DROP DATAFILE statements may be executed in either order.

You can obtain information about data files used by Disk Data tables by querying the FILES table in the INFORMATION_SCHEMA database. An extra “NULL row” was added to this table in MySQL 5.1.14 for providing additional information about undo log files. For more information and examples of use, see Section 22.21, “The INFORMATION_SCHEMA FILES Table”.

Configuration parameters affecting Disk Data behaviour include the following:

  • DiskPageBufferMemory

    This determines the amount of space used for caching pages on disk, and is set in the [NDBD] or [NDBD DEFAULT] section of the config.ini file. It is measured in bytes. Each page takes up 32k. This means that Cluster Disk Data storage always uses N * 32k memory where N is some non-negative integer.

  • SharedGlobalMemory

    This determines the amount of memory that is used for log buffers, disk operations (such as page requests and wait queues), and metadata for tablespaces, log file groups, UNDO files, and data files. It can be set in the [NDBD] or [NDBD DEFAULT] section of the config.ini configuration file, and is measured in bytes.

Note: The OPTIMIZE TABLE statement does not have any effect on Disk Data tables.

15.12. Using High-Speed Interconnects with MySQL Cluster

Even before design of NDB Cluster began in 1996, it was evident that one of the major problems to be encountered in building parallel databases would be communication between the nodes in the network. For this reason, NDB Cluster was designed from the very beginning to allow for the use of a number of different data transport mechanisms. In this Manual, we use the term transporter for these.

The MySQL Cluster codebase includes support for four different transporters:

Most users today employ TCP/IP over Ethernet because it is ubiquitous. TCP/IP is also by far the best-tested transporter for use with MySQL Cluster.

We are working to make sure that communication with the ndbd process is made in “chunks” that are as large as possible because this benefits all types of data transmission.

For users who desire it, it is also possible to use cluster interconnects to enhance performance even further. There are two ways to achieve this: Either a custom transporter can be designed to handle this case, or you can use socket implementations that bypass the TCP/IP stack to one extent or another. We have experimented with both of these techniques using the SCI (Scalable Coherent Interface) technology developed by Dolphin.

15.12.1. Configuring MySQL Cluster to use SCI Sockets

In this section, we show how to adapt a cluster configured for normal TCP/IP communication to use SCI Sockets instead. This documentation is based on SCI Sockets version 2.3.0 as of 01 October 2004.


Any machines with which you wish to use SCI Sockets must be equipped with SCI cards.

It is possible to use SCI Sockets with any version of MySQL Cluster. No special builds are needed because it uses normal socket calls which are already available in MySQL Cluster. However, SCI Sockets are currently supported only on the Linux 2.4 and 2.6 kernels. SCI Transporters have been tested successfully on additional operating systems although we have verified these only with Linux 2.4 to date.

There are essentially four requirements for SCI Sockets:

  • Building the SCI Socket libraries.

  • Installation of the SCI Socket kernel libraries.

  • Installation of one or two configuration files.

  • The SCI Socket kernel library must enabled either for the entire machine or for the shell where the MySQL Cluster processes are started.

This process needs to be repeated for each machine in the cluster where you plan to use SCI Sockets for inter-node communication.

Two packages need to be retrieved to get SCI Sockets working:

  • The source code package containing the DIS support libraries for the SCI Sockets libraries.

  • The source code package for the SCI Socket libraries themselves.

Currently, these are available only in source code format. The latest versions of these packages at the time of this writing were available as (respectively) DIS_GPL_2_5_0_SEP_10_2004.tar.gz and SCI_SOCKET_2_3_0_OKT_01_2004.tar.gz. You should be able to find these (or possibly newer versions) at

Package Installation

Once you have obtained the library packages, the next step is to unpack them into appropriate directories, with the SCI Sockets library unpacked into a directory below the DIS code. Next, you need to build the libraries. This example shows the commands used on Linux/x86 to perform this task:

shell> tar xzf DIS_GPL_2_5_0_SEP_10_2004.tar.gz
shell> cd DIS_GPL_2_5_0_SEP_10_2004/src/
shell> tar xzf ../../SCI_SOCKET_2_3_0_OKT_01_2004.tar.gz
shell> cd ../adm/bin/Linux_pkgs
shell> ./make_PSB_66_release

It is possible to build these libraries for some 64-bit procesors. To build the libraries for Opteron CPUs using the 64-bit extensions, run make_PSB_66_X86_64_release rather than make_PSB_66_release. If the build is made on an Itanium machine, you should use make_PSB_66_IA64_release. The X86-64 variant should work for Intel EM64T architectures but this has not yet (to our knowledge) been tested.

Once the build process is complete, the compiled libraries will be found in a zipped tar file with a name along the lines of DIS-<operating-system>-time-date. It is now time to install the package in the proper place. In this example we will place the installation in /opt/DIS. (Note: You will most likely need to run the following as the system root user.)

shell> cp DIS_Linux_2.4.20-8_181004.tar.gz /opt/
shell> cd /opt
shell> tar xzf DIS_Linux_2.4.20-8_181004.tar.gz
shell> mv DIS_Linux_2.4.20-8_181004 DIS

Network Configuration

Now that all the libraries and binaries are in their proper place, we need to ensure that the SCI cards have proper node IDs within the SCI address space.

It is also necessary to decide on the network structure before proceeding. There are three types of network structures which can be used in this context:

  • A simple one-dimensional ring

  • One or more SCI switches with one ring per switch port

  • A two- or three-dimensional torus.

Each of these topologies has its own method for providing node IDs. We discuss each of them in brief.

A simple ring uses node IDs which are non-zero multiples of 4: 4, 8, 12,...

The next possibility uses SCI switches. An SCI switch has 8 ports, each of which can support a ring. It is necessary to make sure that different rings use different node ID spaces. In a typical configuration, the first port uses node IDs below 64 (4 – 60), the next 64 node IDs (68 – 124) are assigned to the next port, and so on, with node IDs 452 – 508 being assigned to the eighth port.

Two- and three-dimensional torus network structures take into account where each node is located in each dimension, incrementing by 4 for each node in the first dimension, by 64 in the second dimension, and (where applicable) by 1024 in the third dimension. See Dolphin's Web site for more thorough documentation.

In our testing we have used switches, although most large cluster installations use 2- or 3-dimensional torus structures. The advantage provided by switches is that, with dual SCI cards and dual switches, it is possible to build with relative ease a redundant network where the average failover time on the SCI network is on the order of 100 microseconds. This is supported by the SCI transporter in MySQL Cluster and is also under development for the SCI Socket implementation.

Failover for the 2D/3D torus is also possible but requires sending out new routing indexes to all nodes. However, this requires only 100 milliseconds or so to complete and should be acceptable for most high-availability cases.

By placing cluster data nodes properly within the switched architecture, it is possible to use 2 switches to build a structure whereby 16 computers can be interconnected and no single failure can hinder more than one of them. With 32 computers and 2 switches it is possible to configure the cluster in such a manner that no single failure can cause the loss of more than two nodes; in this case, it is also possible to know which pair of nodes is affected. Thus, by placing the two nodes in separate node groups, it is possible to build a “safe” MySQL Cluster installation.

To set the node ID for an SCI card use the following command in the /opt/DIS/sbin directory. In this example, -c 1 refers to the number of the SCI card (this is always 1 if there is only 1 card in the machine); -a 0 refers to adapter 0; and 68 is the node ID:

shell> ./sciconfig -c 1 -a 0 -n 68

If you have multiple SCI cards in the same machine, you can determine which card has which slot by issuing the following command (again we assume that the current working directory is /opt/DIS/sbin):

shell> ./sciconfig -c 1 -gsn

This will give you the SCI card's serial number. Then repeat this procedure with -c 2, and so on, for each card in the machine. Once you have matched each card with a slot, you can set node IDs for all cards.

After the necessary libraries and binaries are installed, and the SCI node IDs are set, the next step is to set up the mapping from hostnames (or IP addresses) to SCI node IDs. This is done in the SCI sockets configuration file, which should be saved as /etc/sci/scisock.conf. In this file, each SCI node ID is mapped through the proper SCI card to the hostname or IP address that it is to communicate with. Here is a very simple example of such a configuration file:

#host           #nodeId
alpha           8
beta            12   16

It is also possible to limit the configuration so that it applies only to a subset of the available ports for these hosts. An additional configuration file /etc/sci/scisock_opt.conf can be used to accomplish this, as shown here:

#-key                        -type        -values
EnablePortsByDefault                yes
EnablePort                  tcp           2200
DisablePort                 tcp           2201
EnablePortRange             tcp           2202 2219
DisablePortRange            tcp           2220 2231

Driver Installation

With the configuration files in place, the drivers can be installed.

First, the low-level drivers and then the SCI socket driver need to be installed:

shell> cd DIS/sbin/
shell> ./drv-install add PSB66
shell> ./scisocket-install add

If desired, the installation can be checked by invoking a script which verifies that all nodes in the SCI socket configuration files are accessible:

shell> cd /opt/DIS/sbin/
shell> ./

If you discover an error and need to change the SCI socket configuration, it is necessary to use ksocketconfig to accomplish this task:

shell> cd /opt/DIS/util
shell> ./ksocketconfig -f

Testing the Setup

To ensure that SCI sockets are actually being used, you can employ the latency_bench test program. Using this utility's server component, clients can connect to the server to test the latency of the connection. Determining whether SCI is enabled should be fairly simple from observing the latency. (Note: Before using latency_bench, it is necessary to set the LD_PRELOAD environment variable as shown later in this section.)

To set up a server, use the following:

shell> cd /opt/DIS/bin/socket
shell> ./latency_bench -server

To run a client, use latency_bench again, except this time with the -client option:

shell> cd /opt/DIS/bin/socket
shell> ./latency_bench -client server_hostname

SCI socket configuration should now be complete and MySQL Cluster ready to use both SCI Sockets and the SCI transporter (see Section, “SCI Transport Connections”).

Starting the Cluster

The next step in the process is to start MySQL Cluster. To enable usage of SCI Sockets it is necessary to set the environment variable LD_PRELOAD before starting ndbd, mysqld, and ndb_mgmd. This variable should point to the kernel library for SCI Sockets.

To start ndbd in a bash shell, do the following:

bash-shell> export LD_PRELOAD=/opt/DIS/lib/
bash-shell> ndbd

In a tcsh environment the same thing can be accomplished with:

tcsh-shell> setenv LD_PRELOAD=/opt/DIS/lib/
tcsh-shell> ndbd

Note: MySQL Cluster can use only the kernel variant of SCI Sockets.

15.12.2. Understanding the Impact of Cluster Interconnects

The ndbd process has a number of simple constructs which are used to access the data in a MySQL Cluster. We have created a very simple benchmark to check the performance of each of these and the effects which various interconnects have on their performance.

There are four access methods:

  • Primary key access

    This is access of a record through its primary key. In the simplest case, only one record is accessed at a time, which means that the full cost of setting up a number of TCP/IP messages and a number of costs for context switching are borne by this single request. In the case where multiple primary key accesses are sent in one batch, those accesses share the cost of setting up the necessary TCP/IP messages and context switches. If the TCP/IP messages are for different destinations, additional TCP/IP messages need to be set up.

  • Unique key access

    Unique key accesses are similar to primary key accesses, except that a unique key access is executed as a read on an index table followed by a primary key access on the table. However, only one request is sent from the MySQL Server, and the read of the index table is handled by ndbd. Such requests also benefit from batching.

  • Full table scan

    When no indexes exist for a lookup on a table, a full table scan is performed. This is sent as a single request to the ndbd process, which then divides the table scan into a set of parallel scans on all cluster ndbd processes. In future versions of MySQL Cluster, an SQL node will be able to filter some of these scans.

  • Range scan using ordered index

    When an ordered index is used, it performs a scan in the same manner as the full table scan, except that it scans only those records which are in the range used by the query transmitted by the MySQL server (SQL node). All partitions are scanned in parallel when all bound index attributes include all attributes in the partitioning key.

To check the base performance of these access methods, we have developed a set of benchmarks. One such benchmark, testReadPerf, tests simple and batched primary and unique key accesses. This benchmark also measures the setup cost of range scans by issuing scans returning a single record. There is also a variant of this benchmark which uses a range scan to fetch a batch of records.

In this way, we can determine the cost of both a single key access and a single record scan access, as well as measure the impact of the communication media used, on base access methods.

In our tests, we ran the base benchmarks for both a normal transporter using TCP/IP sockets and a similar setup using SCI sockets. The figures reported in the following table are for small accesses of 20 records per access. The difference between serial and batched access decreases by a factor of 3 to 4 when using 2KB records instead. SCI Sockets were not tested with 2KB records. Tests were performed on a cluster with 2 data nodes running on 2 dual-CPU machines equipped with AMD MP1900+ processors.

Access TypeTCP/IP SocketsSCI Socket
Serial pk access400 microseconds160 microseconds
Batched pk access28 microseconds22 microseconds
Serial uk access500 microseconds250 microseconds
Batched uk access70 microseconds36 microseconds
Indexed eq-bound1250 microseconds750 microseconds
Index range24 microseconds12 microseconds

We also performed another set of tests to check the performance of SCI Sockets vis-à-vis that of the SCI transporter, and both of these as compared with the TCP/IP transporter. All these tests used primary key accesses either serially and multi-threaded, or multi-threaded and batched.

The tests showed that SCI sockets were about 100% faster than TCP/IP. The SCI transporter was faster in most cases compared to SCI sockets. One notable case occurred with many threads in the test program, which showed that the SCI transporter did not perform very well when used for the mysqld process.

Our overall conclusion was that, for most benchmarks, using SCI sockets improves performance by approximately 100% over TCP/IP, except in rare instances when communication performance is not an issue. This can occur when scan filters make up most of processing time or when very large batches of primary key accesses are achieved. In that case, the CPU processing in the ndbd processes becomes a fairly large part of the overhead.

Using the SCI transporter instead of SCI Sockets is only of interest in communicating between ndbd processes. Using the SCI transporter is also only of interest if a CPU can be dedicated to the ndbd process because the SCI transporter ensures that this process will never go to sleep. It is also important to ensure that the ndbd process priority is set in such a way that the process does not lose priority due to running for an extended period of time, as can be done by locking processes to CPUs in Linux 2.6. If such a configuration is possible, the ndbd process will benefit by 10–70% as compared with using SCI sockets. (The larger figures will be seen when performing updates and probably on parallel scan operations as well.)

There are several other optimized socket implementations for computer clusters, including Myrinet, Gigabit Ethernet, Infiniband and the VIA interface. We have tested MySQL Cluster so far only with SCI sockets. See Section 15.12.1, “Configuring MySQL Cluster to use SCI Sockets” for information on how to set up SCI sockets using ordinary TCP/IP for MySQL Cluster.

15.13. Known Limitations of MySQL Cluster

In this section, we provide a list of known limitations in MySQL Cluster releases in the 5.1.x series compared to features available when using the MyISAM and InnoDB storage engines. Currently, there are no plans to address these in coming releases of MySQL 5.1; however, we will attempt to supply fixes for these issues in subsequent release series. If you check the “Cluster” category in the MySQL bugs database at, you can find known bugs which (if marked “5.1”) we intend to correct in upcoming releases of MySQL 5.1.

The list here is intended to be complete with respect to the conditions just set forth. You can report any discrepancies that you encounter to the MySQL bugs database using the instructions given in Section 1.8, “How to Report Bugs or Problems”. If we do not plan to fix the problem in MySQL 5.1, we will add it to the list.

(Note: See the end of this section for a list of issues in MySQL 5.0 Cluster that have been resolved in the current version.)

Limitations and other issues specific to MySQL Cluster Replication are described in Section 15.10.3, “Known Issues”.

  • Noncompliance in syntax (resulting in errors when running existing applications):

    • Temporary tables are not supported.

    • Text indexes are not supported. That is, you cannot create indexes on columns of any of the TEXT datatypes, nor does the NDB storage engine support FULLTEXT indexes (these are supported by MyISAM only). However, you can index VARCHAR columns of NDB tables.

    • A BIT column cannot be a primary key, unique key, or index, nor can it be part of a composite primary key, unique key, or index.

    • Geometry datatypes (WKT and WKB) are supported in NDB tables in MySQL 5.1. However, spatial indexes are not supported.

    • CREATE TABLE statements may be no more than 4096 characters in length. This limitation affects MySQL 5.1.6, 5.1.7, and 5.1.8 only. (See Bug#17813)

    • In MySQL 5.1.7 and earlier, INSERT IGNORE, UPDATE IGNORE, and REPLACE are supported only for primary keys, but not for unique keys. One possible workaround is to remove the constraint by dropping the unique index, perform any inserts, and then add the unique index again.

      This limitation is removed for INSERT IGNORE and REPLACE in MySQL 5.1.8. (Bug#17431)

    • Support for user-defined partitioning for MySQL Cluster in MySQL 5.1 is restricted to [LINEAR] KEY partitioning. Beginning with MySQL 5.1.12, using any other partitioning type with ENGINE=NDB or ENGINE=NDBCLUSTER in a CREATE TABLE statement results in an error.

      As of MySQL 5.1.6, all Cluster tables are by default partitioned by KEY using the table's primary key as the partitioning key. If no primary key is explicitly set for the table, the “hidden” primary key automatically created by the NDB storage engine is used instead. For additional discussion of these and related issues, see Section 16.2.4, “KEY Partitioning”.

      It is not possible to drop partitions from NDB tables using ALTER TABLE ... DROP PARTITION. The other partitioning extensions to ALTER TABLEADD PARTITION, REORGANIZE PARTITION, and COALESCE PARTITION — are supported for Cluster tables, but use copying and so are not optimised. See Section 16.3.1, “Management of RANGE and LIST Partitions” and Section 13.1.2, “ALTER TABLE Syntax”.

    • When using row-based replication with MySQL Cluster, binary logging cannot be disabled. That is, the NDB storage engine ignores the value of SQL_LOG_BIN. (Bug#16680)

    • The auto_increment_increment and auto_increment_offset server system variables are not supported for Cluster replication.

  • Non-compliance in limits or behavior (may result in errors when running existing applications):

    • Memory Usage:

      Memory comsumed when data is inserted into an NDB table is not automatically recovered when deleted, as it is with other storage engines. Instead, the following rules hold true:

      • A DELETE statement on an NDB table makes the memory formerly used by the deleted rows available for re-use by inserts on the same table only. The memory cannot be used by other NDB tables.

      • A DROP TABLE or TRUNCATE operation on an NDB table frees the memory that was used by this table for re-use by any NDB table — either by the same table or by another NDB table.

        (Recall that TRUNCATE drops and re-creates the table. See Section 13.2.9, “TRUNCATE Syntax”.)

        Memory freed by DELETE operations but still allocated to a specific table can also be made made available for general re-use by performing a rolling restart of the cluster. See Section 15.5.1, “Performing a Rolling Restart of the Cluster”.

    • Error Reporting:

      • A duplicate key error returns the error message ERROR 23000: Can't write; duplicate key in table 'tbl_name'.

      • Like other MySQL storage engines, the NDB storage engine can handle a maximum of one AUTO_INCREMENT column per table. However, in the case of a Cluster table with no explicit primary key, an AUTO_INCREMENT column is automatically defined and used as a “hidden” primary key. For this reason, you cannot define a table that has an explicit AUTO_INCREMENT column unless that column is also declared using the PRIMARY KEY option.

        Attempting to create a table with an AUTO_INCREMENT column that is not the table's primary key, and using the NDB storage engine, fails with an error.

    • Transaction Handling:

      • NDB Cluster supports only the READ COMMITTED transaction isolation level.

      • There is no partial rollback of transactions. A duplicate key or similar error results in a rollback of the entire transaction.

      • Important: If a SELECT from a Cluster table includes a BLOB or TEXT column, the READ COMMITTED transaction isolation level is converted to a read with read lock. This is done to guarantee consistency, due to the fact that parts of the values stored in columns of these types are actually read from a separate table.

      • As noted elsewhere in this chapter, MySQL Cluster does not handle large transactions well; it is better to perform a number of small transactions with a few operations each than to attempt a single large transaction containing a great many operations.

        Among other considerations, large transactions require very large amounts of memory. Because of this, the transactional behaviour of a number of MySQL statements is effected as described in the following list:

        • TRUNCATE is not transactional when used on NDB tables. If a TRUNCATE fails to empty the table, then it must be re-run until it is successful.

        • DELETE FROM (even with no WHERE clause) is transactional. For tables containing a great many rows, you may find that performance is improved by using several DELETE FROM ... LIMIT ... statements to “chunk” the delete operation. If the objective is to empty the table, then you may wish to use TRUNCATE instead.

        • LOAD DATA INFILE is not transactional when used on NDB tables. During such an operation, the NDB engine can and does commit at will.

          LOAD DATA FROM MASTER is not supported in MySQL Cluster.

        • When copying a table as part of an ALTER TABLE, the creation of the copy is non-transactional. (In any case, this operation is rolled back when the copy is deleted.)

      • Node Start, Stop, or Restart:: Starting, stopping, or restarting a node may give rise to temporary errors causing some transactions to fail. These include the following cases:

        • When first starting a node, it is possible that you may see Error 1204 Temporary failure, distribution changed and similar temporary errors.

        • The stopping or failure of any data node can result in a number of different node failure errors. (However, there should be no aborted transactions when performing a planned shutdown of the cluster.)

        In either of these cases, any errors that are generated must be handled within the application. This should be done by retrying the transaction.

    • A number of hard limits exist which are configurable, but available main memory in the cluster sets limits. See the complete list of configuration parameters in Section 15.4.4, “Configuration File”. Most configuration parameters can be upgraded online. These hard limits include:

      • Database memory size and index memory size (DataMemory and IndexMemory, respectively).

        DataMemory is allocated as 32KB pages. As each DataMemory page is used, it is assigned to a specific table; once allocated, this memory cannot be freed except by dropping the table.

        See Section, “Defining Data Nodes”, for further information about DataMemory and IndexMemory.

      • The maximum number of operations that can be performed per transaction is set using the configuration parameters MaxNoOfConcurrentOperations and MaxNoOfLocalOperations. Note that bulk loading, TRUNCATE TABLE, and ALTER TABLE are handled as special cases by running multiple transactions, and so are not subject to this limitation.

      • Different limits related to tables and indexes. For example, the maximum number of ordered indexes per table is determined by MaxNoOfOrderedIndexes.

    • Database names, table names and attribute names cannot be as long in NDB tables as with other table handlers. Attribute names are truncated to 31 characters, and if not unique after truncation give rise to errors. Database names and table names can total a maximum of 122 characters. (That is, the maximum length for an NDB Cluster table name is 122 characters less the number of characters in the name of the database of which that table is a part.)

    • The maximum number of tables in a Cluster database is limited to 20320.

    • In MySQL 5.1.10 and earlier versions, the maximum number of tables having AUTO_INCREMENT columns — including those belonging to hidden primary keys — is 2048.

      This limitation was lifted in MySQL 5.1.11.

    • The maximum number of attributes per table is limited to 128.

    • The maximum permitted size of any one row is 8KB. Note that each BLOB or TEXT column contributes 256 + 8 = 264 bytes towards this total.

    • The maximum number of attributes per key is 32.

  • Unsupported features (do not cause errors, but are not supported or enforced):

    • The foreign key construct is ignored, just as it is in MyISAM tables.

    • Savepoints and rollbacks to savepoints are ignored as in MyISAM.

    • OPTIMIZE operations are not supported.

    • LOAD TABLE ... FROM MASTER is not supported.

  • Performance and limitation-related issues:

    • There are query performance issues due to sequential access to the NDB storage engine; it is also relatively more expensive to do many range scans than it is with either MyISAM or InnoDB.

    • The Records in range statistic is available but not completely tested or officially supported. This may result in non-optimal query plans in some cases. If necessary, you can employ USE INDEX or FORCE INDEX for that purpose.

    • Unique hash indexes created with USING HASH cannot be used for accessing a table if NULL is given as part of the key.

    • SQL_LOG_BIN has no effect on data operations; however, it is supported for schema operations.

      MySQL Cluster cannot produce a binlog for tables having BLOB columns but no primary key.

      Only the following schema operations are logged in a cluster binlog which is not on the mysqld executing the statement:



      • DROP TABLE









  • Missing features:

    • The only supported isolation level is READ COMMITTED. (InnoDB supports READ COMMITTED, READ UNCOMMITTED, REPEATABLE READ, and SERIALIZABLE.) See Section 15.8.5, “Backup Troubleshooting”, for information on how this can affect backup and restore of Cluster databases.

    • No durable commits on disk. Commits are replicated, but there is no guarantee that logs are flushed to disk on commit.

  • Problems relating to multiple MySQL servers (not relating to MyISAM or InnoDB):

    • ALTER TABLE is not fully locking when running multiple MySQL servers (no distributed table lock).

    • DDL operations are not node failure safe. If a node fails while trying to peform one of these (such as CREATE TABLE or ALTER TABLE), the data dictionary is locked and no further DDL statements can be executed without restarting the cluster.

  • Issues exclusive to MySQL Cluster (not related to MyISAM or InnoDB):

    • All machines used in the cluster must have the same architecture. That is, all machines hosting nodes must be either big-endian or little-endian, and you cannot use a mixture of both. For example, you cannot have a management node running on a PowerPC which directs a data node that is running on an x86 machine. This restriction does not apply to machines simply running mysql or other clients that may be accessing the cluster's SQL nodes.

    • Online adding or dropping of nodes is not possible (the cluster must be restarted in such cases).

    • While it is possible to run multiple cluster processes concurrently on a single host, it is not always advisable to do so for reasons of performance and high availability, as well as other considerations. In particular, we do not in MySQL 5.1 support for production use any MySQL Cluster deployment in which more than one ndbd process is run on a single physical machine.

      We may support multiple data nodes per host in a future MySQL release, following additional testing. However, in MySQL 5.1, such configurations can be considered experimental only.

    • Use of Disk Data tables is not supported when running the cluster in diskless mode. Beginning with MySQL 5.1.12, it is disallowed altogether. (Bug#20008)

    • When using multiple management servers:

      • You must give nodes explicit IDs in connectstrings because automatic allocation of node IDs does not work across multiple management servers.

      • You must take extreme care to have the same configurations for all management servers. No special checks for this are performed by the cluster.

    • Multiple network addresses per data node are not supported. Use of these is liable to cause problems: In the event of a data node failure, an SQL node waits for confirmation that the data node went down but never receives it because another route to that data node remains open. This can effectively make the cluster inoperable.

      It is possible to use multiple network hardware interfaces (such as Ethernet cards) for a single data node, but these must be bound to the same address. This also means that it not possible to use more than one [TCP] section per connection in the config.ini file. See Section, “Cluster TCP/IP Connections”, for more information.

    • The maximum number of data nodes is 48.

    • The total maximum number of nodes in a MySQL Cluster is 63. This number includes all SQL nodes (MySQL Servers), API nodes (applications accessing the cluster other than MySQL servers), data nodes, and management servers.

    • The maximum number of metadata objects in MySQL 5.1 Cluster is 20320. This limit is hard-coded.

  • MySQL Cluster issues from previous versions that have been resolved in MySQL 5.1:

    • The NDB Cluster storage engine now supports variable-length column types for in-memory tables.

      Previously, this meant that — for example — any Cluster table having one or more VARCHAR fields which contained only relatively small values, much more memory and disk space were required when using the NDBCluster storage engine than would have been the case for the same table and data using the MyISAM engine. In other words, in the case of a VARCHAR column, such a column required the same amount of storage as a CHAR column of the same size. In MySQL 5.1, this is no longer the case for in-memory tables, where storage requirements for variable-length column types such as VARCHAR and BINARY are comparable to those for these column types when used in MyISAM tables (see Section 11.5, “Data Type Storage Requirements”).

      Important: For MySQL Cluster Disk Data tables, the fixed-width limitation continues to apply. See Section 15.11, “MySQL Cluster Disk Data Storage”.

    • It is now possible to use MySQL replication with Cluster databases. For details, see Section 15.10, “MySQL Cluster Replication”.

      However, circular replication is not currently supported for MySQL Cluster. See Section 15.10.3, “Known Issues”.

    • Autodiscovery of databases is now supported for multiple MySQL servers accessing the same MySQL Cluster, provided that a given mysqld is already running and is connected to the cluster at the time that the database is created on a different mysqld.

      What this means is that if a mysqld process first connects to the cluster after a database named db_name has been created, you should issue a CREATE SCHEMA db_name statement on the “new” MySQL server when it first accesses that MySQL Cluster. Once this has been done, the “newmysqld should be able to detect any tables in that database tables without errors.

      This also means that online schema changes in NDB tables are now possible. That is, the result of operations such as ALTER TABLE and CREATE INDEX performed on one SQL node in the cluster are now visible to the cluster's other SQL nodes without any additional action being taken.

    • Beginning with MySQL 5.1.10, it is possible to perform a Cluster backup and restore between different architectures. Previously — for example — you could not back up a cluster running on a big-endian platform and then restore from that backup to a cluster running on a little-endian system. (Bug#19255)

    • Beginning with MySQL 5.1.10, it is possible to install MySQL with Cluster support to a non-default location and change the search path for font description files using either the --basedir or --character-sets-dir options. (Previously, ndbd in MySQL 5.1 searched only the default path — typically /usr/local/mysql/share/mysql/charsets — for character sets.)

    • In MySQL 5.1, it is no longer necessary, when running multiple management servers, to restart all the cluster's data nodes to enable the management nodes to see one another.

15.14. MySQL Cluster Development Roadmap

In this section, we discuss changes in the implementation of MySQL Cluster in MySQL 5.1 as compared to MySQL 5.0.

There are a number of significant changes in the implementation of the NDB Cluster storage engine in MySQL 5.1 as compared to that in MySQL 5.0. For an overview of these changes, see Section 15.14.1, “MySQL Cluster Changes in MySQL 5.1”

15.14.1. MySQL Cluster Changes in MySQL 5.1

A number of new features for MySQL Cluster have been implemented in MySQL 5.1:

  • Integration of MySQL Cluster into MySQL replication: This makes it possible to update from any MySQL Server in the cluster and still have the MySQL Replication handled by one of the MySQL Servers in the cluster, with the state of the slave side remaining consistent with the cluster acting as the master.

    See Section 15.10, “MySQL Cluster Replication”.

  • Support for disk-based records: Records on disk are now supported. Indexed fields including the primary key hash index must still be stored in RAM but all other fields can be on disk.

    See Section 15.11, “MySQL Cluster Disk Data Storage”.

  • Variable-sized records: A column defined as VARCHAR(255) currently uses 260 bytes of storage independent of what is stored in any particular record. In MySQL 5.1 Cluster tables, only the portion of the column actually taken up by the record will be stored. This will make possible a reduction in space requirements for such columns by a factor of 5 in many cases.

  • User-defined partitioning: Users can define partitions based on columns that are part of the primary key. It is possible to partition NDB tables based on KEY and LINEAR KEY schemes. This feature is also available for many other MySQL storage engines, which support additional partitioning types that are not available with NDB Cluster tables.

    For additional general information about user-defined partitioning in MySQL 5.1, see Chapter 16, Partitioning. Specifics of partitioning types are discussed in Section 16.2, “Partition Types”.

    The MySQL Server can also determine whether it is possible to “prune away” some of the partitions from the WHERE clause. See Section 16.4, “Partition Pruning”.

  • Autodiscovery of table schema changes. In MySQL 5.1, you no longer need to issue FLUSH TABLES or a “dummySELECT in order for new NDB tables or changes made to schemas of existing NDB tables on one SQL node to be visible on the cluster's other SQL nodes.

    Note: When creating a new database, it is still necessary to issue the CREATE DATABASE or CREATE SCHEMA statement on each SQL node in the cluster.

See MySQL Cluster issues from previous versions that have been resolved in MySQL 5.1 for more information.

15.15. MySQL Cluster Glossary

The following terms are useful to an understanding of MySQL Cluster or have specialized meanings when used in relation to it.

  • Cluster:

    In its generic sense, a cluster is a set of computers functioning as a unit and working together to accomplish a single task.

    NDB Cluster:

    This is the storage engine used in MySQL to implement data storage, retrieval, and management distributed among several computers.

    MySQL Cluster:

    This refers to a group of computers working together using the NDB storage engine to support a distributed MySQL database in a shared-nothing architecture using in-memory storage.

  • Configuration files:

    Text files containing directives and information regarding the cluster, its hosts, and its nodes. These are read by the cluster's management nodes when the cluster is started. See Section 15.4.4, “Configuration File”, for details.

  • Backup:

    A complete copy of all cluster data, transactions and logs, saved to disk or other long-term storage.

  • Restore:

    Returning the cluster to a previous state, as stored in a backup.

  • Checkpoint:

    Generally speaking, when data is saved to disk, it is said that a checkpoint has been reached. More specific to Cluster, it is a point in time where all committed transactions are stored on disk. With regard to the NDB storage engine, there are two types of checkpoints which work together to ensure that a consistent view of the cluster's data is maintained:

    • Local Checkpoint (LCP):

      This is a checkpoint that is specific to a single node; however, LCP's take place for all nodes in the cluster more or less concurrently. An LCP involves saving all of a node's data to disk, and so usually occurs every few minutes. The precise interval varies, and depends upon the amount of data stored by the node, the level of cluster activity, and other factors.

    • Global Checkpoint (GCP):

      A GCP occurs every few seconds, when transactions for all nodes are synchronized and the redo-log is flushed to disk.

  • Cluster host:

    A computer making up part of a MySQL Cluster. A cluster has both a physical structure and a logical structure. Physically, the cluster consists of a number of computers, known as cluster hosts (or more simply as hosts. See also Node and Node group below.

  • Node:

    This refers to a logical or functional unit of MySQL Cluster, and is sometimes also referred to as a cluster node. In the context of MySQL Cluster, we use the term “node” to indicate a process rather than a physical component of the cluster. There are three node types required to implement a working MySQL Cluster:

    • Management (MGM) nodes:

      Manages the other nodes within the MySQL Cluster. It provides configuration data to the other nodes; starts and stops nodes; handles network partitioning; creates backups and restores from them, and so forth.

    • SQL (MySQL server) nodes:

      Instances of MySQL Server which serve as front ends to data kept in the cluster's data nodes. Clients desiring to store, retrieve, or update data can access an SQL node just as they would any other MySQL Server, employing the usual authentication methods and API's; the underlying distribution of data between node groups is transparent to users and applications. SQL nodes access the cluster's databases as a whole without regard to the data's distribution across different data nodes or cluster hosts.

    • Data nodes:

      These nodes store the actual data. Table data fragments are stored in a set of node groups; each node group stores a different subset of the table data. Each of the nodes making up a node group stores a replica of the fragment for which that node group is responsible. Currently, a single cluster can support up to 48 data nodes total.

    It is possible for more than one node to co-exist on a single machine. (In fact, it is even possible to set up a complete cluster on one machine, although one would almost certainly not want to do this in a production environment.) It may be helpful to remember that, when working with MySQL Cluster, the term host refers to a physical component of the cluster whereas a node is a logical or functional component (that is, a process).

    Note Regarding Terms: In older versions of the MySQL Cluster documentation, data nodes were sometimes referred to as “database nodes”. The term “storage nodes” has also been used. In addition, SQL nodes were sometimes known as “client nodes”. This older terminology has been deprecated to minimize confusion, and for this reason should be avoided. They are also often referred to as “API nodes” — an SQL node is actually an API node that provides an SQL interface to the cluster.

  • Node group:

    A set of data nodes. All data nodes in a node group contain the same data (fragments), and all nodes in a single group should reside on different hosts. It is possible to control which nodes belong to which node groups.

    For more information, see Section 15.2.1, “MySQL Cluster Nodes, Node Groups, Replicas, and Partitions”.

  • Node failure:

    MySQL Cluster is not solely dependent upon the functioning of any single node making up the cluster; the cluster can continue to run if one or more nodes fail. The precise number of node failures that a given cluster can tolerate depends upon the number of nodes and the cluster's configuration.

  • Node restart:

    The process of restarting a failed cluster node.

  • Initial node restart:

    The process of starting a cluster node with its filesystem removed. This is sometimes used in the course of software upgrades and in other special circumstances.

  • System crash (or system failure):

    This can occur when so many cluster nodes have failed that the cluster's state can no longer be guaranteed.

  • System restart:

    The process of restarting the cluster and reinitializing its state from disk logs and checkpoints. This is required after either a planned or an unplanned shutdown of the cluster.

  • Fragment:

    A portion of a database table; in the NDB storage engine, a table is broken up into and stored as a number of fragments. A fragment is sometimes also called a “partition”; however, “fragment” is the preferred term. Tables are fragmented in MySQL Cluster in order to facilitate load balancing between machines and nodes.

  • Replica:

    Under the NDB storage engine, each table fragment has number of replicas stored on other data nodes in order to provide redundancy. Currently, there may be up four replicas per fragment.

  • Transporter:

    A protocol providing data transfer between nodes. MySQL Cluster currently supports four different types of transporter connections:

    • TCP/IP

      This is, of course, the familiar network protocol that underlies HTTP, FTP (and so on) on the Internet. TCP/IP can be used for both local and remote connections.

    • SCI

      Scalable Coherent Interface is a high-speed protocol used in building multiprocessor systems and parallel-processing applications. Use of SCI with MySQL Cluster requires specialized hardware, as discussed in Section 15.12.1, “Configuring MySQL Cluster to use SCI Sockets”. For a basic introduction to SCI, see this essay at

    • SHM

      Unix-style shared memory segments. Where supported, SHM is used automatically to connect nodes running on the same host. The Unix man page for shmop(2) is a good place to begin obtaining additional information about this topic.

    Note: The cluster transporter is internal to the cluster. Applications using MySQL Cluster communicate with SQL nodes just as they do with any other version of MySQL Server (via TCP/IP, or through the use of Unix socket files or Windows named pipes). Queries can be sent and results retrieved using the standard MySQL client APIs.

  • NDB:

    This stands for Network Database, and refers to the storage engine used to enable MySQL Cluster. The NDB storage engine supports all the usual MySQL data types and SQL statements, and is ACID-compliant. This engine also provides full support for transactions (commits and rollbacks).

  • shared-nothing architecture:

    The ideal architecture for a MySQL Cluster. In a true shared-nothing setup, each node runs on a separate host. The advantage such an arrangement is that there no single host or node can act as single point of failure or as a performance bottle neck for the system as a whole.

  • In-memory storage:

    All data stored in each data node is kept in memory on the node's host computer. For each data node in the cluster, you must have available an amount of RAM equal to the size of the database times the number of replicas, divided by the number of data nodes. Thus, if the database takes up 1GB of memory, and you want to set up the cluster with four replicas and eight data nodes, a minimum of 500MB memory will be required per node. Note that this is in addition to any requirements for the operating system and any other applications that might be running on the host.

    In MySQL 5.1, it is also possible to create Disk Data tables where non-indexed columns are stored on disk, thus reducing the memory footprint required by the cluster. Note that indexes and indexed column data are still stored in RAM. See Section 15.11, “MySQL Cluster Disk Data Storage”.

  • Table:

    As is usual in the context of a relational database, the term “table” denotes a set of identically structured records. In MySQL Cluster, a database table is stored in a data node as a set of fragments, each of which is replicated on additional data nodes. The set of data nodes replicating the same fragment or set of fragments is referred to as a node group.

  • Cluster programs:

    These are command-line programs used in running, configuring, and administering MySQL Cluster. They include both server daemons:

    • ndbd:

      The data node daemon (runs a data node process)

    • ndb_mgmd:

      The management server daemon (runs a management server process)

    and client programs:

    • ndb_mgm:

      The management client (provides an interface for executing management commands)

    • ndb_waiter:

      Used to verify status of all nodes in a cluster

    • ndb_restore:

      Restores cluster data from backup

    For more about these programs and their uses, see Section 15.6, “Process Management in MySQL Cluster”.

  • Event log:

    MySQL Cluster logs events by category (startup, shutdown, errors, checkpoints, and so on), priority, and severity. A complete listing of all reportable events may be found in Section 15.7.3, “Event Reports Generated in MySQL Cluster”. Event logs are of two types:

    • Cluster log:

      Keeps a record of all desired reportable events for the cluster as a whole.

    • Node log:

      A separate log which is also kept for each individual node.

    Under normal circumstances, it is necessary and sufficient to keep and examine only the cluster log. The node logs need be consulted only for application development and debugging purposes.