Oracle GlassFish Server Message Queue Developer's Guide for C Clients Release 4.5.2 Part Number E24944-01 |
|
|
View PDF |
This chapter describes how to use C functions to accomplish specific tasks and provides brief code samples to illustrate some of these tasks. (For clarity, the code examples shown in the following sections omit a function call status check.)
Following a brief discussion of overall design and a summary of client tasks, the topics covered include the following:
This chapter does not provide exhaustive information about each function. For detailed function information, please see the description of that function in Reference.
For information on building Message Queue C programs, see Client Design Issues.
The general procedures for producing and consuming messages are introduced below. The procedures have a number of common steps which need not be duplicated if a client is both producing and consuming messages.
Call the MQCreateProperties
function to get a handle to a properties object.
Use one or more of the MQSet...Property
functions to set connection properties that specify the name of the broker, its port number, and its behavior.
Use the MQCreateConnection
function to create a connection.
Use the MQCreateSession
function to create a session and to specify its acknowledge mode and its receive mode. If the session will be used only for producing messages, use the receive mode MQ_SESSION_SYNC_RECEIVE
to avoid creating a thread for asynchronous message delivery.
Use the MQCreateDestination
function to specify a physical destination on the broker. The destination name you specify must be the same as the name of the physical destination.
Use the MQCreateMessageProducer
function or the MQCreateMessageProducerForDestination
function to create a message producer. (If you plan to send a lot of messages to the same destination, you should use the MQCreateMessageProducerForDestination
function.)
Use the MQCreateBytesMessage
function or the MQCreateTextMessage
function to get a newly created message handle.
Call the MQCreateProperties
function to get a handle to a properties object that will describe the message header properties. This is only required if you want to set a message header property.
Use one or more of the MQSet...Property
functions to set properties that specify the value of the message header properties you want to set.
Use the MQSetMessageHeaders
function, passing a handle to the properties object you created in Step 8 and Step 9.
Repeat Step 8 if you want to define custom message properties, and then use the MQSetMessageProperties
function to set these properties for your message.
Use the MQSetMessageReplyTo
function if you want to specify a destination where replies to the message are to be sent.
Use one of the MQSendMessage...
functions to send the message.
Call the MQCreateProperties
function to get a handle to a properties object.
Use one or more of the MQSet...Property
functions to set connection properties that specify the name of the broker, its port number, and its behavior.
Use the MQCreateConnection
function to create a connection.
Use the MQCreateSession
function to create a session and to specify its receive mode. Specify MQ_SESSION_SYNC_RECEIVE
for a synchronous session.
Use the MQCreateDestination
function to specify a destination on the broker from which the consumer is to receive messages. The destination name you specify must be the same as the name of the physical destination.
Use the MQCreateMessageConsumer
function or the MQCreateDurableMessageConsumer
function to create a consumer.
Use the MQStartConnection
function to start the connection.
Use one of the MQReceiveMessage...
functions to start message delivery.
Call the MQCreateProperties
function to get a handle to a properties object.
Use one or more of the MQSet...Property
functions to set connection properties that specify the name of the broker, its port number, and its behavior.
Use the MQCreateConnection
function to create a connection.
Use the MQCreateSession
function to create a session and to specify its acknowledge mode and its receive mode. Specify MQ_SESSION_ASYNC_RECEIVE
for asynchronous message delivery.
Use the MQCreateDestination
function to specify a destination on the broker from which the consumer is to receive messages. The logical destination name you specify must be the same as the name of the physical destination.
Write a callback function of type MQMessageListenerFunc
that will be called when the broker starts message delivery. In the body of this callback function, use the functions described in Processing a Message , to process the contents of the incoming message.
Use the MQCreateAsyncMessageConsumer
function or the MQCreateAsyncDurableMessageConsumer
function to create a consumer.
Use the MQStartConnection
function to start the connection and message delivery.
When you create a connection, set message header properties, or set user-defined message properties, you must pass a handle to a properties object. You use the MQCreateProperties
function to create this object and to obtain a handle to it. When you receive a message, you can use specific MQGet...Property
functions to obtain the type and value of each message property.
This section describes the functions you use to set and get properties. A property is defined as a key-value pair.
You use the functions listed in Table 2-1 to create a handle to a properties object, and to set properties. You can use these functions to create and define properties for connections or for individual messages.
Set message properties and message header properties using the same procedure you use to set connection properties. You can set the following message header properties for sending a message:
MQ_CORRELATION_ID_HEADER_PROPERTY
MQ_MESSAGE_TYPE_HEADER_PROPERTY
For more information, see the description of the MQSetMessageProperties function.
Table 2-1 Functions Used to Set Properties
Function | Description |
---|---|
Creates a properties object and passes back a handle to it. |
|
Sets an |
|
Sets an |
|
Sets an |
|
Sets an |
|
Sets an |
|
Sets an |
|
Sets an |
|
Sets an |
Call the MQCreateProperties
function to get a handle to a newly created properties object.
Call one of the MQSet...Property
functions to set one of the connection properties described in Table 4-2 . At a minimum, you must specify the name of the host of the broker to which you want to connect and its port number.
Which function you call depends on the type of the property you want to set; for example, to set an MQString
property, you call the MQSetStringProperty
function; to set an MQBool
property, you call the MQSetBoolProperty
function; and so on. Each function that sets a property requires that you pass a key name and value; these are listed and described in Table 4-2.
When you have set all the properties you want to define for the connection, you can then create the connection, by calling the MQCreateConnection
function.
Once the connection is created with the properties you specify, you cannot change its properties. If you need to change connection properties after you have created a connection, you will need to destroy the old connection and its associated objects and create a new one with the desired properties. It is a good idea to think through the desired behavior before you create a connection.
The code sample below illustrates how you create a properties handle and how you use it for setting connection properties.
MQStatus status; MQPropertiesHandle propertiesHandle = MQ_INVALID_HANDLE; status = (MQCreateProperties(&propertiesHandle); status = (MQSetStringProperty(propertiesHandle, MQ_BROKER_HOST_PROPERTY, "localhost")); status = (MQSetInt32Property(propertiesHandle, MQ_BROKER_PORT_PROPERTY, 7676)); status = MQSetStringProperty(propertiesHandle, MQ_CONNECTION_TYPE_PROPERTY, "TCP"));
The Message Queue C client runtime sets the connection properties that specify the name and version of the Message Queue product; you can retrieve these using the functionMQGetMetaData. These properties are described at the end of Table 4-2, starting with MQ_NAME_PROPERTY
.
When you receive a message, if you are interested in the message properties, you need to obtain a handle to the properties object associated with that message:
Use the MQGetMessageProperties
function to obtain a handle to the properties object for user-defined properties.
If you are interested in any message header properties, use the MQGetMessageHeaderProperties
function to obtain a handle to the header properties. See MQGetMessageHeaders.
Having obtained the handle, you can iterate through the properties and then use the appropriate MQGet...Property
function to determine the type and value of each property.
Table 2-2 lists the functions you use to iterate through a properties handle and to obtain the type and value of each property.
Table 2-2 Functions Used to Get Message Properties
Function | Description |
---|---|
Starts the iteration process through the specified properties handle. |
|
Returns |
|
Passes back the address of the next property key in the referenced property handle. |
|
Gets the type of the specified property. |
|
Gets the value of the specified |
|
Gets the value of the specified |
|
Gets the value of the specified |
|
Gets the value of the specified |
|
Gets the value of the specified |
|
Gets the value of the specified |
|
Gets the value of the specified |
|
Gets the value of the specified |
Start the process by calling the MQPropertiesKeyIterationStart function.
Loop using the MQPropertiesKeyIterationHasNext function.
Extract the name of each property key by calling the MQPropertiesKeyIterationGetNext function.
Determine the type of the property value for a given key by calling the MQGetPropertyType function.
Use the appropriate MQGet...Property
function to find the value of the specified property key and type.
If you know the property key, you can just use the appropriate MQGet...Property
function to get its value. The code sample below illustrates how you implement these steps.
MQStatus status; MQPropertiesHandle headersHandle = MQ_INVALID_HANDLE; MQBool redelivered; ConstMQString my_msgtype; status = (MQGetMessageHeaders(messageHandle, &headersHandle)); status = (MQGetBoolProperty(headersHandle, MQ_REDELIVERED_HEADER_PROPERTY, &redelivered)); status = MQGetStringProperty(headersHandle, MQ_MESSAGE_TYPE_HEADER_TYPE_PROPERTY, &my_msgtype);
All messaging occurs within the context of a connection: the behavior of the connection is defined by the properties set for that connection. You use the functions listed in Table 2-3 to create, start, stop, and close a connection.
Table 2-3 Functions Used to Work with Connections
Function | Description |
---|---|
Initializes the SSL library. You must call this function before you create any connection that uses SSL. |
|
Creates a connection and passes back a handle to it. |
|
Starts the specified connection and starts or resumes delivery of messages. |
|
Stops the specified connection. |
|
Returns a handle to name and version information for the Message Queue product. |
|
Closes the specified connection. |
Before you create a connection, you must do the following:
Define the connection properties. See Setting Connection and Message Properties for more information.
Specify a user name and password for the connection. See "User Authentication" in Oracle GlassFish Server Message Queue Administration Guide for information on how to set up users.
Write a connection exception listener function. You will need to pass a reference to this listener when you create the connection. This function will be called synchronously when a connection exception occurs for this connection. For more information, see Callback Type for Connection Exception Handling .
If you want a secure connection, call the MQIntitializeSSL
function to initialize the SSL library. See Working With Secure Connections for more information.
When you have completed these steps, you are ready to call MQCreateConnection
to create a connection. After you create the connection, you can create a session as described in Working With Sessions and Destinations .
When you send a message, you do not need to start the connection explicitly by calling MQStartConnection
. You do need to call MQStartConnection before the broker can deliver messages to a consumer.
If you need to halt delivery in the course of processing messages, you can call the MQStopConnection function.
Connection properties specify the following information:
The host name and port of the broker to which you want to connect
The transport protocol of the connection service used by the client
How broker and client acknowledgements are handled to support messaging reliability
How message flow is to be managed
How secure messaging should be implemented
The following sections examine the effect of properties used to manage connection handling, reliability, message flow, and security.
Table 4-2 lists and describes all properties of a connection. For information on how to set and change connection properties, see Working With Properties.
Connections to a message server are specified by a broker host name and port number.
Set MQ_BROKER_NAME_PROPERTY
to specify the broker name.
Set MQ_BROKER_PORT_PROPERTY
to specify the port of the broker's port mapper service. In this case, the port mapper will dynamically assign the port to which the client connects.
Set MQ_BROKER_SERVICE_PORT_PROPERTY
to specify the number of a port to which the client connects. This is a static, fixed port assignment; it bypasses the broker's port mapper service. If you do need to connect to a fixed port on the broker, make sure that the service needed is enabled and available at the specified port by setting the imq.serviceName.protocolType.port
broker property.
Set the connection property MQ_CONNECTION_TYPE_PROPERTY
to specify the underlying transport protocol. Possible values are TCP
or SSL
.
Remember that you need to configure the JMS service port on the broker side as well. For example, if you want to connect your client via ssljms
to port 1756, you would do the following.
On the client side: Set the MQ_SERVICE_PORT_PROPERTY
to 1756 and set the MQ_CONNECTION_TYPE_PROPERTY
to SSL
.
On the broker side: Set the imq.serviceNameType.protocol.port property to 1756 as follows.
imq.ssljms.ssl.port=1756
The MQ_PING_INTERVAL_PROPERTY
also affects connection handling. This property is set to the interval (in seconds) that the connection can be idle before the C client runtime pings the broker to test whether the connection is still alive. This property is useful for either producers who use the connection infrequently or for clients who are exclusive consumers, passively waiting for messages to arrive. The default value is 30 seconds. Setting an interval that is too low may result in some performance loss. The minimum permitted value is 1 second to prevent this from happening.
Currently, the C-API does not support auto-reconnect or failover, which allows the client runtime to automatically reconnect to a broker if a connection fails.
Two connection properties enable the acknowledgement of messages sent to the broker and of messages received from the broker. These are described in Message Acknowledgement. In addition to setting these properties, you can also set MQ_ACK_TIMEOUT_PROPERTY
, which determines the maximum time that the client runtime will wait for any broker acknowledgement before throwing an exception.
A number of connection properties determine the use of Message Queue control messages by the client runtime. Messages sent and received by Message Queue clients and Message Queue control messages pass over the same client-broker connection. Because of this, delays may occur in the delivery of control messages, such as broker acknowledgements, if these are held up by the delivery of JMS messages. To prevent this type of congestion, Message Queue meters the flow of JMS messages across a connection.
Set MQ_CONNECTION_FLOW_COUNT_PROPERTY
to specify the number of Message Queue messages in a metered batch. When this number of messages is delivered to the client runtime, delivery is temporarily suspended, allowing any control messages that had been held up to be delivered. Message delivery is resumed upon notification by the client runtime, and continues until the count is again reached.
MQ_CONNECTION_FLOW_LIMIT_PROPERTY
specifies the maximum number of unconsumed messages that can be delivered to a client runtime. When the number of messages reaches this limit, delivery stops and resumes only when the number of unconsumed messages drops below the specified limit. This helps a consuming client that is taking a long time to process messages from being overwhelmed with pending messages that might cause it to run out of memory.
MQ_CONNECTION_FLOW_LIMIT_ENABLED_PROPERTY
specifies whether the value MQ_CONNECTION_FLOW_LIMIT_PROPERTY
is used to control message flow.
You should keep the value of MQ_CONNECTION_FLOW_COUNT_PROPERTY
low if the client is doing operations that require many responses from the broker; for example, the client is using the CLIENT_ACKNOWLEDGE or AUTO_ACKNOWLEDGE
modes, persistent messages, transactions, or if the client is adding or removing consumers. You can increase the value of MQ_CONNECTION_FLOW_COUNT_PROPERTY
without compromising performance if the client has only simple consumers on a connection using DUPS_OK
mode.
The C API does not currently support consumer-based flow control.
Establishing a secure connection between the client and the broker requires both the administrator and the developer to do some additional work. The administrator's work is described in "Message Encryption" in Oracle GlassFish Server Message Queue Administration Guide. In brief, it requires that the administrator do the following:
Generate certificates (self-signed or signed by a certificate authority) and add those certificates to the broker's keystore
Enable the ssljms
connection service in the broker
Provide the password to the certificate keystore when starting the broker
The developer must also do some work to configure the client for secure messaging. The work required depends on whether the broker is trusted (the default setting) and on whether the developer wants to provide an additional means of verification if the broker is not trusted and the initial attempt to create a secure connection fails.
The MessageQueue C-API library uses NSS to support the SSL transport protocol between the Message Queue C client and the Message Queue broker. The developer must take care if the client application using secure Message Queue connections uses NSS (for other purposes) directly as well and does NSS initialization. For additional information, see Coordinating NSS Initialization.
By default the MQ_SSL_BROKER_IS_TRUSTED
property is set to true
, and this means that the Message Queue client runtime will accept any certificate that is presented to it. The following procedure explains what you must do to establish a secure connection.
Follow this procedure:
Set the MQ_CONNECTION_TYPE_PROPERTY
to SSL
.
If you want the runtime to check the broker's certificate, set the MQ_SSL_BROKER_IS_TRUSTED
property to false
. Otherwise, you can leave it to its default (true
) value.
Generate the NSS files cert
N.db
, key
N.db
, and secmod.db
using the certificate database tool certutil
.
You can find this tool in mqInstallHome/nss/bin
.
For directions and an example of using this tool, see
http://www.mozilla.org/projects/security/pki/nss/tools/certutil.html
Note the path name of the directory that contains the NSS files you generated in Configuring the Client for Secure Communication.
If you have set the MQ_SSL_BROKER_IS_TRUSTED
property to false
, use the certutil
tool to import the root certificate of the authority certifying the broker into the database files you generated in Configuring the Client for Secure Communication .
Make sure that the MQ_BROKER_HOST_PROPERTY
value is set to the same value as the (CN) common name in the broker's certificate.
If you have set the MQ_SSL_BROKER_IS_TRUSTED
property to false
, you have the option of enabling broker fingerprint-based verification in case authorization fails. For details, see Verification Using Fingerprints.
Call the function MQInitializeSSL
once (and only once) before creating the connection, and pass the name of the directory that contains the NSS files you generated in Configuring the Client for Secure Communication. If the broker is trusted, these files can be empty.
You must call this function before you create any connection to the broker, including connections that do not use SSL.
If certificate authorization fails when the broker is using a certificate authority, it is possible to give the client runtime another means of establishing a secure connection by comparing broker certificate fingerprints. If the fingerprints match, the connection is granted; if they do not match, the attempt to create the connection will fail.
Follow this procedure:
Set the broker connection property MQ_SSL_CHECK_BROKER_FINGERPRINT
to true
.
Retrieve the broker's certificate fingerprint by using the java keytool -list
option on the broker's keystore file:
You will use the output of this command as the value for the connection property MQ_SSL_BROKER_CERT_FINGERPRINT
in Verification Using Fingerprints. For example, if the output contains a value like the following:
Certificate fingerprint (MD5): F6:A5:C1:F2:E6:63:40:73:97:64:39:6C:1B:35:0F:8E
You would specify this value for MQ_SSL_BROKER_CERT_FINGEPRINT
.
Set the connection property MQ_SSL_BROKER_CERT_FINGEPRINT
to the value obtained in Verification Using Fingerprints.
If your application uses NSS directly, other than to support Message Queue secure communication, you need to coordinate NSS initialization with the Message Queue C-API library. There are two cases to consider:
Your application does not use secure Message Queue connections.
In this case, you should do your application's NSS initialization before calling MQCreateConnection
to create any connection to the Message Queue broker.
Your application does use secure Message Queue connections.
In this case, you should follow the procedure outlined below before calling MQCreateConnection
to create any Message Queue connection.
Follow this procedure:
Call the function MQInitializeSSL.
(You must specify the path to the directory containing the NSS files as the certdbpath
parameter to this function.)
Your application's use of NSS must specify the same certdbpath
value for the location of its NSS files. (That is, the certificates needed by your application must be located in the same directory as the certificates needed by Message Queue.)
Internally, the function MQInitializeSSL
does the following:
Calls the function NSS_Init(certdbpath)
.
Sets DOMESTIC
cipher policy using the function NSS_SetDomesticPolicy()
.
Enables all cipher suites, including RSA_NULL_MD5
by calling the function SSL_CipherPrefSetDefault(SSL_RSA_WITH_NULL_MD5, PR_TRUE)
.
Calls the function SSL_ClearSessionCache()
.
If your application needs different cipher suite settings, after you call the MQInitializeSSL()
function, you can modify the cipher suites by calling the function SSL_CipherPrefSetDefault
. However, note that these changes will affect your secure connection to the Message Queue broker as well.
In order to do an orderly shutdown, you need to close the connection by calling MQCloseConnection and then to free the memory associated with the connection by calling the MQFreeConnection function.
Closing the connection closes all sessions, producers, and consumers created from this connection. This also forces all threads associated with this connection that are blocking in the library to return.
After all the application threads associated with this connection and its descendant sessions, producers, and consumers have returned, the application can call the MQFreeConnection function to release all resources associated with the connection.
To get information about a connection, call the MQGetMetaData function. This returns name and version information for the Message Queue product.
A session is a single-threaded context for producing and consuming messages. You can create multiple producers and consumers for a session, but you are restricted to using them serially. In effect, only a single logical thread of control can use them. A session supports reliable delivery through acknowledgment options or by using transactions.
Table 2-4 describes the functions you use to create and manage sessions.
Table 2-4 Functions Used to Work with Sessions
Function | Description |
---|---|
Creates the specified session and passes back a handle to it. |
|
Passes back the acknowledgement mode of the specified session. |
|
Stops message delivery and restarts message delivery with the oldest unacknowledged message. (For non-transacted sessions.) |
|
Rolls back a transaction associated with the specified session. |
|
Commits a transaction associated with the specified session. |
|
Closes the specified session. |
The MQCreateSession
function creates a new session and initializes a handle to it in the sessionHandle
parameter. The number of sessions you can create for a single connection is limited only by system resources. You can create a session after you have created a connection.
When you create a session, you specify whether it is transacted, the acknowledge mode, and the receive mode. After you create a session, you can create the producers, consumers, and destinations that use the session context to do their work.
If you specify that a session be transacted, the acknowledge mode is ignored. Within a transacted session, the broker tracks sends and receives, completing these operations only when the client issues a call to commit the transaction. If a send or receive operation fails, the operation will return an error. Your application can handle the error by ignoring it, retrying it, or rolling back the entire transaction. When a transaction is committed, all the successful operations are completed. When a transaction is rolled back, all the successful operations are cancelled. A transaction cannot encompass both the production and consumption of the same message.
The scope of a local transaction is a single session. One or more producer or consumer operations can be grouped into a single local transaction only if performed in the context of a single session.
To extend the scope of a transaction beyond a single session, you can use a distributed transaction. A distributed transaction is managed by an external distributed transaction manager, as described in Working With Distributed Transactions.
Both messages that are sent and messages that are received can be acknowledged.
In the case of message producers, if you want the broker to acknowledge its having received a non-persistent message (to its physical destination), you must set the connection's MQ_ACK_ON_PRODUCE_PROPERTY
to MQ_TRUE
. If you do so, the sending function will return only after the broker has acknowledged receipt of the message. By default, the broker acknowledges receipt of persistent messages.
Acknowledgements on the consuming side means that the client runtime acknowledges delivery and consumption of all messages from a physical destination before the message service deletes the message from that destination. You can specify one of the following acknowledge modes for the consuming session when you create that session.
MQ_AUTO_ACKNOWLEDGE
specifies that the session automatically acknowledge each message consumed by the client.
MQ_CLIENT_ACKNOWLEDGE
specifies that the client must explicitly acknowledge messages by calling MQAcknowledgeMessages
. In this case, all messages are acknowledged that have been consumed up to the point where the acknowledge function is called. (This could include messages consumed asynchronously by many different message listeners in that session, independent of the order in which they were consumed.)
MQ_DUPS_OK_ACKNOWLEDGE
specifies that the session acknowledges receipt of messages after each ten messages are consumed. It does not guarantee that messages are delivered and consumed only once.
(The setting of the connection property MQ_ACK_ON_ACKNOWLEDGE_PROPERTY
also determines the effect of some of these acknowledge modes. For more information, see Table 4-2.)
Note:
In the DUPS_OK_ACKNOWLEDGE
mode, the session does not wait for broker acknowledgements. This option can be used in Message Queue C clients for which duplicate messages are not a problem. Also, you can call the MQRecoverSession function to explicitly request redelivery of messages that have been received but not yet acknowledged by the client. When redelivering such messages, the broker will set the header field MQ_REDLIEVERED_HEADER_PROPERTY
.
You can specify a session's receive mode as either MQ_SESSION_SYNC_RECEIVE
or MQ_SESSION_ASYNC_RECEIVE
. If the session you create will be used for sending messages only, you should specify MQ_SESSION_SYNC_RECEIVE
for its receive mode for optimization because the asynchronous receive mode automatically allocates an additional thread for the delivery of messages it expects to receive.
Managing a session involves using threads appropriately for the type of session (synchronous or asynchronous) and managing message delivery for both transacted and nontransacted sessions. For more information about thread management, see Managing Threads.
For a session that is not transacted, use the MQRecoverSession function to restart message delivery with the last unacknowledged message.
For a session that is transacted, use the MQRollBackSession function to roll back any messages that were delivered within this transaction. Use the MQCommitSession function to commit all messages associated with this transaction.
Use the MQCloseSession function to close a session and all its associated producers and consumers. This function also frees memory allocated for the session.
You can get information about a session's acknowledgment mode by calling the MQGetAcknowledgeMode function.
After creating a session, you can create destinations or temporary destinations for the messages you want to send. Table 2-5 lists the functions you use to create and to get information about destinations.
Table 2-5 Functions Used to Work with Destinations
Functions | Description |
---|---|
Creates a destination and initializes a handle to it. |
|
Creates a temporary destination and initializes a handle to it. |
|
Returns the type (queue or topic) of the specified destination. |
A destination refers to where a message is destined to go. A physical destination is a JMS message service entity (a location on the broker) to which producers send messages and from which consumers receive messages. The message service provides the routing and delivery for messages sent to a physical destination.
When a Message Queue C client creates a destination programmatically using the MQCreateDestination
function, a destination name must be specified. The function initializes a handle to a destination data type that holds the identity (name) of the destination. The important thing to remember is that this function does not create the physical destination on the broker; this must be done by the administrator. The destination that is created programmatically however must have the exact same name and type as the physical destination created on the broker. For example, if you use the MQCreateDestination
function to create a queue destination called myMailQDest
, the administrator has to create a physical destination on the broker named myMailQDest
.
Destination names starting with "mq" are reserved and should not be used by client programs.
When you create a destination, you must also specify its type: MQ_QUEUE_DESTINATION
or MQ_TOPIC_DESTINATION
. See "Messaging Domains" in Oracle GlassFish Server Message Queue Technical Overview for a discussion of these two types of destinations and how to choose the type that suits your needs.
By default, the imq.autocreate.topic
and imq.autocreate.queue
broker properties are turned on. In this case, which is more convenient in a development environment, the broker automatically creates a physical destination whenever a message consumer or message producer attempts to access a non-existent destination. The auto-created physical destination will have the same name as that of the destination you created using the MQCreateDestination
function.
You use the MQCreateTemporaryDestination
function to create a temporary destination. You can use such a destination to implement a simple request/reply mechanism. When you pass the handle of a temporary destination to the MQSetMessageReplyTo
function, the consumer of the message can use that handle as the destination to which it sends a reply.
Temporary destinations are explicitly created by client applications and are automatically deleted when the connection is closed. They are maintained (and named) by the broker only for the duration of the connection for which they are created. Temporary destinations are system-generated uniquely for their connection and only their own connection is allowed to create message consumers for them.
Use the MQGetDestinationType
function to determine the type of a destination: queue or topic. There may be times when you do not know the type of the destination to which you are replying: for example, when you get a handle from the MQGetMessageReplyTo
function. Because the semantics of queue and topic destinations differ, you need to determine the type of a destination in order to reply appropriately.
This section describes how you use the C-API to complete the following tasks:
Compose a message
Send a message
Receive a message
Process a message
You can create either a text message or a bytes message. A message, whether text or bytes, is composed of a header, properties, and a body. You can also create a message type which has no body.
Table 2-6 lists the functions you use to construct messages.
Table 2-6 Functions Used to Construct Messages
Function | Description |
---|---|
Creates an |
|
Creates an |
|
Creates an |
|
Sets message header properties. (Optional) |
|
Sets user-defined message properties. |
|
Sets the body of an |
|
Sets the body of an |
|
Specifies the destination where replies to this message should be sent. |
A header is required of every message. Header fields contain values used for routing and identifying messages.
Some header field values are set automatically by Message Queue during the process of producing and delivering a message, some depend on settings specified when message producers send a message, and others are set on a message-by-message basis by the client using the MQSetMessageHeader
function. Table 2-7 lists the header fields defined (and required) by JMS and their corresponding names, as defined by the C-API.
Table 2-7 JMS-defined Message Header
JMS Message Header Field | C-API Message Header Property Name |
---|---|
|
Defined implicitly when a producer sends a message to a destination, or when a consumer receives a message from a destination. |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Set by the |
|
|
For additional information about each property type and who sets it, see MQSetMessageHeaders.
JMS specifies six classes (or types) of messages. The C-API supports only three of these types, as described in Table 2-8. If a Message Queue C client expects to receive messages from a Message Queue Java client, it will be unable to process messages whose body types are other than those described in the table. It will also be unable to process messages that are compressed by the Message Queue Java client runtime.
Table 2-8 C-API Message Body Types
Type | Description |
---|---|
|
A message whose body contains an |
|
A message whose body contains a stream of uninterpreted bytes. |
|
A message consisting of a header and (optional) properties, but no body. |
Create a message using either the MQCreateBytesMessage function or the MQCreateTextMessage function. Either of these functions returns a message handle that you can then pass to the functions you use to set the message body, header, and properties (listed in Composing Messages). You can also use the MQCreateMessage function to create a message that has a header and properties but no message body.
Use the MQSetTextMessageText function to define the body of a text message; use the MQSetBytesMessageBytes function to define the body of a bytes message.
Use the MQSetMessageHeaders to set any message header properties.
The message header can specify up to eight properties; most of these are set by the client runtime when sending the message or are set by the broker. The client can set MQ_CORRELATION_ID_HEADER_PROPERTY
and MQ_MESSAGE_TYPE_HEADER_PROPERTY
for sending a message.
Use the MQSetMessageProperties function to set any user-defined properties for this message.
When you set message header properties or when you set additional user-defined properties, you must pass a handle to a properties object that you have created using the MQCreateProperties function. For more information, see Working With Properties.
You can use the MQSetMessageReplyTo function to associate a message with a destination that recipients can use for replies. To do this, you must first create a destination that will serve as your reply-to destination. Then, pass a handle to that destination when you call the MQSetMessageReplyTo function. The receiver of a message can use the MQGetMessageReplyTo function to determine whether a sender has set up a destination where replies are to be sent.
Messages are sent by a message producer within the context of a connection and a session. Once you have obtained a connection, created a session, and composed your message, you can use the functions listed in Table 2-9to create a message producer and to send the message.
Which function you choose to send a message depends on the following factors:
Whether you want the send function to override certain message header properties
Send functions whose names end in Ext
allow you to override default values for priority, time-to-live, and delivery mode header properties.
Whether you want to send the message to the destination associated with the message producer
If you created a message producer with no specified destination, you must used one of the ...ToDestination
send functions. If you created a message producer with a specified destination, you must use one of the other send functions.
Table 2-9 Functions for Sending Messages
Function | Action |
---|---|
Creates a message producer with no specified destination. |
|
Creates a message producer with a specified destination. |
|
Sends a message for the specified producer. |
|
Sends a message for the specified producer and allows you to set priority, time-to-live, and delivery mode. |
|
Sends a message to the specified destination. |
|
Sends a message to the specified destination and allows you to set priority, time-to-live, and delivery mode. |
If you send a message using one of the functions that does not allow you to override header properties, the following message header fields are set to default values by the send function.
MQ_PERSISTENT_HEADER_PROPERTY
will be set to MQ_PERSISTENT_DELIVERY
.
MQ_PRIORITY_HEADER_PROPERTY
will be set to 4.
MQ_EXPIRATION_HEADER_PROPERTY
will be set to 0, which means that the message will never expire.
To override these values, use one of the extended send functions. For a complete list of message header properties, see MQGetMessageHeaders .
Message headers also contain fields that can be set by the sending client; in addition, you can set user-defined message properties as well. For more information, see Composing Messages.
You can set the connection property MQ_ACK_ON_PRODUCE_PROPERTY
when you create the connection to make sure that the message has reached its destination on the broker:
By default, the broker acknowledges receiving persistent messages only.
If you set the property to MQ_TRUE
, the broker acknowledges receipt of all messages (persistent and non-persistent) from the producing client.
If you set the property to MQ_FALSE
, the broker does not acknowledge receipt of any message (persistent or non-persistent) from the producing client.
Note that "acknowledgement" in this case is not programmatic but internally implemented. That is, the client thread is blocked and does not return until the broker acknowledges messages it receives.
An administrator can set a broker limit, REJECT_NEWEST
, which allows the broker to avert memory problems by rejecting the newest incoming message. If the incoming message is persistent, then an error is returned which the sending client should handle, perhaps by retrying the send a bit later. If the incoming message is not persistent, the client has no way of knowing that the broker rejected it. The broker might also reject a message if it exceeds a specified limit.
Messages are received by a message consumer in the context of a connection and a session. In order to receive messages, you must explicitly start the connection by calling the MQStartConnection
function.
Table 2-10 lists the functions you use to create message consumers and to receive messages.
Table 2-10 Functions Used to Receive Messages
Function | Description |
---|---|
Creates the specified synchronous consumer and passes back a handle to it. |
|
Creates a durable synchronous message consumer for the specified destination. |
|
Creates an asynchronous message consumer for the specified destination. |
|
Creates a durable asynchronous message consumer for the specified destination. |
|
Unsubscribes the specified durable message consumer. |
|
Passes a handle back to a message delivered to the specified consumer if a message is available; otherwise it returns an error. |
|
Passes a handle back to a message delivered to the specified consumer if a message is available; otherwise it blocks until a message becomes available. |
|
Passes a handle back to a message delivered to the specified consumer if a message is available within the specified amount of time. |
|
Acknowledges the specified message and all messages received before it on the same session |
|
Closes the specified consumer. |
When you create a consumer, you need to make several decisions:
Do you want to receive messages synchronously or asynchronously?
If you create a synchronous consumer, you can call one of three kinds of receive functions to receive your messages. If you create an asynchronous consumer, you must specify the name of a callback function that the client runtime can call when a message is delivered to the destination for that consumer. For information about the callback function signature, see Callback Type for Asynchronous Message Consumption .
If you are consuming messages from a topic, do you want to use a durable or a nondurable consumer?
A durable consumer receives all the messages published to a topic, including the ones published while the subscriber is inactive. A nondurable consumer only receives messages while the subscriber is active.
The broker retains a record of this durable subscription and makes sure that all messages from the publishers to this topic are retained until they are either acknowledged by this durable subscriber or until they have expired. Sessions with durable subscribers must always provide the same client identifier. In addition, each consumer must specify a durable name using the durableName
parameter, which uniquely identifies (for each client identifier) each durable subscription it creates.
A session's consumers are automatically closed when you close the session or connection to which they belong. However, messages will be routed to the durable subscriber while it is inactive and delivered when a new durable consumer is recreated. To close a consumer without closing the session or connection to which it belongs, use the MQCloseMessageConsumer function. If you want to close a durable consumer permanently, you should call the function MQUnsubscribeDurableMessageConsumer after closing it, to delete state information maintained by the broker on behalf of the durable consumer.
If you have created a synchronous consumer, you can use one of three receive functions: MQReceiveMessageNoWait
, MQReceiveMessageWait
, or MQReceiveMessagewithTimeOut
. In order to use any of these functions, you must have specified MQ_SESSION_SYNC_RECEIVE
for the receive mode when you created the session.
When you create a session you must specify one of several acknowledge modes for that session. If you specify MQ_CLIENT_ACKNOWLEDGE
as the acknowledge mode for the session, you must explicitly call the MQAcknowledgeMessages
function to acknowledge messages that you have received. If the session is transacted, the acknowledge mode parameter is ignored.
When the receiving function returns, it gives you a handle to the delivered message. You can pass that handle to the functions described in Processing a Message, in order to read message properties and information stored in the header and body of the message.
It is possible that a message can be lost for synchronous consumers in a session using AUTO_ACKNOWLEDGE
mode if the provider fails. To prevent this possibility, you should either use a transacted session or a session in CLIENT_ACKNOWLEDGE
mode.
Because distributed applications involve greater processing time, such an application might not behave as expected if it were run locally. For example, calling the MQReceiveMessageNoWait
function might return MQ_NO_MESSAGE
even when there is a message available to be retrieved on the broker. See the usage notes provided in the section MQReceiveMessageNoWait for more information.
To receive a message asynchronously, you must create an asynchronous message consumer and pass the name of an MQMessageListenerFunc
type callback function. (Therefore, you must set up the callback function before you create the asynchronous consumer that will use it.) You should start the connection only after creating an asynchronous consumer. If the connection is already started, you should stop the connection before creating an asynchronous consumer.
You are also responsible for writing the message listener function. Mainly, the function needs to process the incoming message by examining its header, body, and properties, or it needs to pass control to a function that can do this processing. The client is also responsible for freeing the message handle (either from within the listener or from outside of the listener) by calling the MQFreeMessage
function.
When you create a session you must specify one of several acknowledge modes for that session. If you specify MQ_CLIENT_ACKNOWLEDGE
as the acknowledge mode for the session, you must explicitly call the MQAcknowledgeMessages
function to acknowledge messages that you have received.
For more information about the signature and content of a call back function, see Callback Type for Asynchronous Message Consumption.
When the callback function is called by the session delivery of a message, it gives you a handle to the delivered message. You can pass that handle to the functions described in Processing a Message, in order to read message properties and information stored in the header and body of the message.
When a message is delivered to you, you can examine the message's properties, type, headers, and body. The functions used to process a message are described in Processing a Message.
Table 2-11 Functions Used to Process Messages
Function | Description |
---|---|
Gets message header properties. |
|
Gets user-defined message properties. |
|
Gets the message type: |
|
Gets the body of an |
|
Gets the body of an |
|
Gets the destination where replies to this message should be sent. |
If you are interested in a message's header information, you need to call the MQGetMessageHeaders
function. If you need to read or check any user-defined properties, you need to call the MQGetMessageProperties
function. Each of these functions passes back a properties handle. For information on how you can read property values, see Getting Message Properties.
Before you can examine the message body, you can call the MQGetMessageType
function to determine whether the message is a text or bytes message. You can then call the MQGetTextMessageText
, or the MQGetBytesMessageBytes
function to get the contents of the message.
Some message senders specify a reply destination for their message. Use the MQGetMessageReplyTo
function to determine that destination.
In accordance with the X/Open distributed transaction model, Message Queue C-API support for distributed transactions relies upon a distributed transaction manager. The distributed transaction manage tracks and manages distributed transactions, coordinating the decision to commit them or roll them back, and coordinating failure recovery. The Message Queue C-API supports the X/Open XA interface, qualifying it as an XA-compliant resource manager. This support allows C-API clients running in a distributed transaction processing environment to participate in distributed transactions.
In particular, two C-API functions support the participation of C-API clients in distributed transactions:
MQGetXAConnection() MQCreateXASession()
If a C-client application is to be used in the context of a distributed transaction, then it must obtain a connection by using MQGetXAConnection()
and create a session for producing and consuming messages by using MQCreateXASession()
. The start, commit, and rollback, of any distributed transaction is managed by the distributed transaction manager.
For more information on XA resource managers, see the XA Specification.
In accordance with the X/Open XA interface specification, a distributed transaction manager needs the following information regarding the Message Queue XA-compliant resource manager:
Name of the xa_switch_t
structure: sun_my_xa_switch
Name of the Resource Manager: SUN_RM
The MQ C-API library to be linked: mqcrt
The xa_close
string and format: none
The xa_open
string and format: semicolon (";")-separated name/value pairs
The following name/value pairs are supported:
Table 2-12 Message Queue Resource Manager Name/Value Pairs
Name | Value | Description | Default |
---|---|---|---|
address |
host:port |
The host:port of the broker's Portmapper service. |
|
username |
string |
The username for connecting to the broker |
|
password |
string |
The username's password |
|
conntype |
|
The protocol type of the connection to the broker |
|
trustedhost |
|
Whether the broker host is trusted (only applicable for |
|
certdbpath |
string |
The full path to the directory that contains NSS certificate and key database files |
not set |
clientid |
string |
Required only for JMS durable subscriptions |
not set |
reconnects |
integer |
The number of re-connection attempts to broker (0 means no reconnect) |
|
To help you program an application that uses distributed transactions, Message Queue provides programming examples based on the Tuxedo distributed transaction manager. A description of the sample programs and their location is provided in Table 1-3.
Nearly all Message Queue C functions return an MQStatus
result. You can use this return value to determine whether the function returned successfully and, if not, to determine the cause of the error.
Table 2-13 lists the functions you use to get error information.
Table 2-13 Functions Used in Handling Errors
Function | Description |
---|---|
Returns an |
|
Returns the error code for the specified |
|
Returns a descriptive string for the specified |
|
Returns the calling thread's current error trace or |
Call MQStatusIsError
, passing it an MQStatus
result for the function whose result you want to test.
If the MQStatusIsError
function returns MQ_TRUE
, call MQGetStatusCode
or MQGetStatusString
to identify the error.
If the status code and string information is not sufficient to identify the cause of the error, you can get additional diagnostic information by calling MQGetErrorTrace
to obtain the calling thread's current error trace if this information is available.
Reference lists common errors returned for each function. In addition to these errors, the following error codes may be returned by any Message Queue C function:
MQ_STATUS_INVALID_HANDLE
MQ_OUT_OF_MEMORY
MQ_NULL_PTR_ARG
In addition, the MQ_TIMEOUT_EXPIRED
can return from any Message Queue C function that communicates with the Message Queue broker if the connection MQ_ACK_TIMEOUT_PROPERTY
is set to a non-zero value.
Table 2-14 lists the functions you use to free or deallocate memory allocated by the Message Queue-C client library on behalf of the user. Such deallocation is part of normal memory management and will prevent memory leaks.
The functions MQCloseConnection
, MQCloseSession
, MQCloseMessageProducer
, and MQCloseMessageConsumer
are used to free resources associated with connections, sessions, producers, and consumers.
Table 2-14 Functions Used to Free Memory
Function | Description |
---|---|
Frees memory allocated to the specified connection. |
|
Frees memory allocated to the specified destination. |
|
Frees memory allocated to the specified message. |
|
Frees memory allocated to the specified properties handle. |
|
Frees memory allocated to the specified |
You should free a connection only after you have closed the connection with the MQCloseConnection function and after all of the application threads associated with this connection and its dependent sessions, producers, and consumers have returned.
You should not free a connection while an application thread is active in a library function associated with this connection or one of its dependent sessions, producers, consumers, and destinations.
Freeing a connection does not release resources held by a message associated with this connection. You must free memory allocated for this message by explicitly calling the MQFreeMessage
function.
You should not free a properties handle if the properties handle passed to a function becomes invalid on its return. If you do, you will get an error.
The Message Queue C-API library uses two environment variables to control execution-time logging:
MQ_LOG_FILE
specifies the file to which log messages are directed. If you do not specify a file name for this variable, stderr
is used. If MQ_LOG_FILE
is a directory name, it should include a trailing directory separator.
By default, .n (where n is 0, 1, 2,...) is appended to the actual log file name. This is used as a rotation index, and the indices are used sequentially when the maximum log file size is reached. You can use %g
to specify a rotation index replacement in MQ_LOG_FILE
after the last directory separator. Only the last %g
is used if multiple %g's
are specified. the %g
replacement can be escaped with %
. The maximum rotation index is 9, and the maximum log file size is 1 MB. These limits are not configurable.
MQ_LOG_LEVEL
specifies a numeric level that indicates the detail of logging information needed:
-1
OFF, which suppresses the generation of log messages
0
SEVERE level, which indicates a severe failure
1
WARNING level, which indicates a potential problem
2
INFO level, for informational messages
3
CONFIG level, for static configuration messages
4
FINE level, which provides tracing information
5
FINER level, which provides detailed tracing information
6
FINEST level, which provides highly detailed tracing information
For any value of MQ_LOG_LEVEL
, log messages of that value or lower are generated. For example, setting MQ_LOG_LEVEL
to 2
causes INFO, WARNING and SEVERE log messages to be generated.
The default level is 3
, CONFIG.