Working with NFS and iSCSI Filesystem Images

Filesystem Images

You must create one or more filesystem images to run on your QRM nodes.  Filesystem Images live on your NFS or iSCSI server.  We must define this server in openQRM by creating a Storage Server using the GUI or via the command line.  Please see the Storage Server creation instructions for further details on how to do this.  

Note that after you create the definition of the image within openQRM, that changes to this will not affect the image in storage.  For example, if you clone the image definition, you still need to clone the storage data on the server.  These definitions are like pointers that openQRM and the managed resources will use to connect to the central storage during the provisioning life cycle.

Creating a NFS Filesystem Image

openQRM can capture filesystem images from running linux machines in your network. The easiest way to create your first filesystem image is to capture the filesystem on the openQRM server itself and put it in NFS storage.  The time it takes to create a filesystem image creation depends on multiple factors.  For a relatively simple filesystem image less than three Gigabytes in size, allow about one hour.

To create a NFS filesystem image from your openQRM server, follow these steps:

  1. Be sure you have created a openQRM Storage Server using the GUI or CLI.
  2. Create the filesystem image using the CLI
    1. Run 'qrm-filesystem-image create' for help using CLI options.
    2. For example, you might run  'qrm-filesystem-image create --filesystem-image redhatAS3U6 -l localhost:/ -t mynfsserver:/fs-images/redhatAS3U6'
Some qrm-filesystem-image CLI options explained:

Note that qrm-filesystem-image is only used to create NFS filesystem images.

    Network interfaces

If the system from which you are creating the new application image has network interfaces configured, you may wish to configure new interfaces for the new image, or you may prefer to keep these configurations.

    Services

You may wish to configure new services for the new image, or you may prefer to keep the services that are configured for the source file system. For example, if httpd is running on the server from which you are creating the application image, but you do not need httpd in the new application image, you might want to create the new image without services.

 
Create the image from a file system on a remote machine

You can create a new filesystem image from filesystems running on another server on the network.  To do so, change the -l flag from qrm-cli to give the ip address and path from the remote server.  We then use the CLI similar to before to create the filesystem image.



Creating an iSCSI Filesystem Image

openQRM has support for two different types of iSCSI implementations.  The first is software iSCSI initiated, and the second is hardware initiated iSCSI.   More information about the software initiated iSCSI implementation can be found here.  More information on the hardware initiated iSCSI can be found below in the installation section.  iSCSI images can either be created through the automatic filesystem creation process upon a reboot of the source system, or in the case of a hardware HBA by doing a direct installation to the iSCSI LUN via the HBA.  

The following steps talk about the client configuration needed for iSCSI, but doesn't mention the server implementation details since openQRM does not ship with an iSCSI target.  You should be able to use any target that conforms to the iSCSI specifications whether it is hardware, software, opensource or commercial.  Some of these implementations will require some extra steps to manage the server configuration.

Automatic filesystem creation

Some additional options for automatically creating a filesystem  exist for NFS and iSCSI (software initiated).  Please see the instructions for NFS and iSCSI automatic image creation for more details.

iSCSI HBA (hardware initiated) installation

If you have a iSCSI HBA card, you can install to the iSCSI LUN directly as if it's a local drive.  When you use the HBA card, the operating system sees the iSCSI drive as a local so it is no different than installing an operating system like normal (e.g. depending on the OS the drive will look like /dev/sda or C:).  After the OS installation, you'll need to install the openQRM agent into the image.  The following outlines these steps.
  1. Create a Storage server definition with the iScsi_hw type.
  2. Configure the HBA to connect to the iSCSI target you have configured to serve these nodes.  This may involve changing the BIOS of the HBA to point to the IP and LUN you wish to connect to.
  3. Download the drivers for your OS if they are not included in the distribution, and put them on some type of installation media (like a floppy disk).
  4. Reboot the system and start the installation
  5. Both Windows and Linux should prompt you for the option to install extra drivers.  You should load the drivers downloaded in step 2, and then complete the OS installation.
  6. Reboot the system into the OS and login and install the agent software for your OS (detailed instructions below)
  7. Shutdown the installation node.

Instructions for installing the openQRM agent for Windows

  1. For windows, you'll need to install the openqrm-win-plugin*rpm on the openQRM server, and this will give you the agent installation in a file $QRM_BASEDIR/qrm/plugins/win/install/qrminfod-win32.exe.  
  2. Copy this to the windows installation system, and run the executable.
  3. Define the "Destination Folder" and click next.
  4. Enter the "QRM server IP Address" only.  This should be the IP address of the openQRM server that is installed on your network.  Click Next.
  5. Define "QRM username & password" and press Install.

Instructions for installing the openQRM agent for Linux
  1. Install the agent rpm openqrm-core-resource*.rpm  directly onto the installation system
  2. Edit the file /opt/qrm/resource/etc/qrm-bootstrap.conf and change the QRMIP to point to the openQRM server that is installed on your network.



Filesystem Management

Viewing Available filesystem Images
You can view the filesystems images that exist on the openQRM server in the GUI management console.

Cloning a Filesystem Image
After you create an application image, you can create a duplicate copy of it for other Virtual Environments to use.  You can then modify this filesystem image for your specific needs by mounting it directly and chrooting into it, or by assigning it to a new Virtual Environment.

To copy a filesystem image, follow these steps:

  1. Clone the filesystem image definition within openQRM with the qrm-cli command.  The following is an example:

    qrm-cli filesystem clone --name <existing filesystem image name> --new_name clone1 --identifier /openqrm-filesystem-images/webserver1

  2. On your NFS server, manually copy the directory of the original filesystem image to a directory with the new filesystem image name.  The following is an example:

    cp -a /openqrm-filesystem-images/redhatAS3U6  /openqrm-filesystem-images/webserver1


Renaming a Filesystem Image

After you create a filesystem image, you can use the management console or the command line interface to rename it.

To rename a filesystem image, follow these steps:

  1. Stop any Virtual Environments that are running the filesystem image that is to be renamed.  
  2. From the command line on the openQRM server use qrm-cli filesystem rename.  For example:

    ./qrm-cli filesystem rename --name test-webserver --new_name production-webserver

  3. On the NFS server, rename the directory containing the filesystem image to its new name using the `mv` command.

  4. On each Virtual Environment, reassign the application image by its new name.

  5. Restart each Virtual Environment.


Deleting a Filesystem Image

You can remove a filesystem image from your QRM server using the management console or the command line interface.

To delete an filesystem image using the CLI, follow these steps:

  1. Stop all Virtual Environments in which the filesystem image is running.
  2. From the command line on the openQRM server use qrm-cli filesystem rename.  For example:

    ./qrm-cli filesystem delete --name webserver3

  3. On your NFS server, manually delete the directory containing your filesystem image.


Testing Application Image Changes
The user can test a modification to an existing environment without disrupting a running production VE and have the ability to easily roll back the changes. To do so:
  1. Clone the existing environment using techniques described above.
  2. Optionally, you may want to create and use a new kernel image with your new filesystem image.  See the user manual for how to work with kernel images.
  3. Create a new Virtual Environment for testing based on the newly-created filesystem image (and optionally the new kernel image), and deploy it to a new node.
  4. Log into that Node - the IP address is shown in the node status menu - and modify it to your needs.
  5. When the new environment works properly stop your production VE and modify it to use the new filesystem image (and optionally the kernel image).  Restart the VE.
  6. If it is desirable to roll back the changes, you can modify the VE to its original settings.