8 Message and Transactional Inflow

This chapter describes how WebLogic resource adapters use inbound connections to handle message inflow and transactional inflow.

This chapter includes the following sections:

Overview of Message and Transactional Inflow

Message inflow refers to inbound communication from an EIS to the application server, using a resource adapter. Inbound messages can be part of a transaction that is governed by a Transaction Manager that is external to WebLogic Server and the resource adapter, as described in Transactional Inflow.

The following diagram provides an overview of how messaging and transaction inflow occurs within a resource adapter and the role played by the Work Manager.

Figure 8-1 Messaging and Transactional Inflow Architecture

Description of Figure 8-1 follows
Description of "Figure 8-1 Messaging and Transactional Inflow Architecture"

Architecture Components

Figure 8-1 contains the following components:

  • A client application, which connects to an application running on WebLogic Server, but which also needs to connect to an EIS

  • An external system (in this case, an EIS or Enterprise Information System)

  • An application component (an EJB) that the client application uses to submit outbound requests to the EIS through the resource adapter

  • A message endpoint application (a message-driven bean and possibly other Java EE components) used for the receipt of inbound messages from the EIS through the resource adapter

  • The WebLogic Server Work Manager and an associated thread (or threads) to which the resource adapter submits Work instances to process inbound messages and possibly process other actions.

  • An external Transaction Manager, to which the WebLogic Server Transaction Manager is subordinate for transactional inflow of messages from the EIS

  • The WebLogic Server Connector container in which the resource adapter is deployed. The container manages the following:

    • A deployed resource adapter that provides bi-directional (inbound and outbound) communication to and from the EIS.

    • An active Work instance.

    • Multiple managed connections (MC1, ..., MCn), which are objects representing the outbound physical connections from the resource adapter to the EIS.

    • Connection handles (C-handle) returned to the application component from the connection factory of the resource adapter and used by the application component for communicating with the EIS.

    • One of perhaps many activation specifications. There is an activation specification (ActivationSpec) that corresponds to each specific message listener type, MLT-j. For information about requirements for an ActivationSpec class, see Chapter 13, "Message Inflow" in JSR 322: Java EE Connector Architecture 1.6.

    • One of the connection pools maintained by the container for the management of managed connections for a given ManagedConnectionFactory (in this case, MCF-2. A Connector container could include multiple connection pools, each corresponding to a different type of connections to a single EIS or even different EISes).

    • A MessageEndpointFactory created by the EJB container and used by the resource adapter to create proxies to MessageEndpoint instances (MDB instances from the MDB pool).

  • An external message source, which could be an EIS or Message Provider

Inbound Communication Scenario

This section describes a basic inbound communication scenario that may be described using the diagram, showing how inbound messages originate in an EIS, flow into the resource adapter, and are handled by a Message-driven Bean. For related information, see Figure 2-1.

A typical simplified inbound sequence involves the following steps:

  1. The EIS sends a message to the resource adapter.

  2. The resource adapter inspects the message and determines what type of message it is.

  3. The resource adapter may create a Work object and submit it to the Work Manager. The Work Manager performs the succeeding work in a separate Thread, while the resource adapter can continue waiting for other incoming messages.

  4. Based on the message type, the resource adapter (either directly or as part of a Work instance) looks up the correct message endpoint to which it will send the message.

  5. Using the message endpoint factory corresponding to the type of message endpoint it needs, the resource adapter creates a message endpoint (which is a proxy to a message-driven bean instance from the MDB pool).

  6. The resource adapter invokes the message listener method on the endpoint, passing it message content based on the message it received from the EIS.

  7. The message is handled by the MDB in one of several possible ways:

    1. the MDB may handle the message directly and possibly return a result to the EIS through the resource adapter

    2. the MDB may distribute the message to some other application component

    3. the MDB may place the message on a queue to be picked up by the client

    4. the MDB may directly communicate with the client application.

How Message Inflow Works

A resource adapter that supports inbound communication from an EIS to the application server typically includes the following:

  • A proprietary communications channel and protocol for connecting to and communicating with an EIS. The communications channel and protocol are not visible to the application server in which the resource adapter is deployed. See Appendix 8, "Proprietary Communications Channel and Protocol."

  • One or more message types recognized by the resource adapter.

  • A dispatching mechanism to dispatch a message of a given type to another component in the application server.

Handling Inbound Messages

A resource adapter may handle an inbound message in a variety of ways. For example, it may:

  • Handle the message locally, that is, within the ResourceAdapter bean, without involving other components.

  • Pass the message off to another application component. For example, it may look up an EJB and invoke a method on it.

  • Send the message to a message endpoint. Typically, a message endpoint is a message-driven bean (MDB). For more information, see Appendix 8, "Message Inflow to Message Endpoints (Message-driven Beans)."

Inbound messages may return a result to the EIS that is sending the message. A message requiring an immediate response is referred to as synchronous (the sending system waits for a response). This is also referred to as request-response messaging. A message that does not expect a response as part of the same exchange with the resource adapter is referred to as asynchronous or event notification-based communication. A resource adapter can support asynchronous or synchronous communications for all three destinations listed above.

Depending upon the transactional capabilities of the resource adapter and the EIS, inbound messages can be either part of a transaction (XA) or not (non-transactional). If the messages are XA, the controlling transaction may be coordinated by an external Transaction Manager (transaction inflow) or by the application server's Transaction Manager. See Appendix 8, "Transactional Inflow."

In most cases, inbound messages in a resource adapter are dispatched through a Work instance in a separate thread. The resource adapter wraps the work to be done in a Work instance and submits it to the application server's Work Manager for execution and management. A resource adapter can submit a Work instance using the doWork(), startWork(), or scheduleWork() methods depending upon the scheduling requirements of the work.

Proprietary Communications Channel and Protocol

The resource adapter can expose connection configuration information to the deployer by various means; for example, as properties on the ResourceAdapter bean or properties on the ActivationSpec object. An alternative is to use the same communication channel for inbound as well as outbound traffic. Thus you can also set configuration information on the outbound connection pool.

Message Inflow to Message Endpoints (Message-driven Beans)

Prior to EJB 2.1, a message-driven bean (MDB) supported only Java Message Service (JMS) messaging. That is, an MDB had to implement the javax.jms.MessageListener interface, including the onMessage(javax.jms.Message) message listener method. MDBs were bound to JMS components and the JMS subsystem delivered the messages to MDBs by invoking the onMessage() method on an instance of the MDB.

With EJB 2.1, the JMS-only MDB restriction has been lifted to accommodate the delivery of messages from inbound resource adapters. The main ingredients for message delivery to an MDB by way of a resource adapter are:

  • An inbound message of a certain type (determined by the resource adapter/EIS contract)

  • An ActivationSpec object implemented by the resource adapter

  • A mapping between message types and message listener interfaces

  • An MDB that implements a given message listener interface

  • A deployment-time binding between an MDB and a resource adapter

For more information about message-driven Beans, see "Message-Driven EJBs" in Programming Enterprise JavaBeans for Oracle WebLogic Server.

Deployment-Time Binding Between an MDB and a Resource Adapter

A resource adapter can be deployed independently (as a standalone RAR) or as part of an enterprise application (EAR). An MDB can also be deployed independently (as a standalone JAR) or as part of an enterprise application (EAR). In either case, an MDB whose messages are derived from a resource adapter must be bound to the resource adapter. The following sections describe binding the MDB and resource adapter and subsequent messaging operations.

Binding an MDB and a Resource Adapter

To bind an MDB and a resource adapter, you must:

  1. Set the jndi-name element in the weblogic-ra.xml deployment descriptor for the resource adapter. See jndi-name in Appendix A, "weblogic-ra.xml Schema."

  2. Set the adapter-jndi-name element in the weblogic-ejb-jar.xml deployment descriptor to match the value set in the corresponding jndi-name element in the resource adapter.

  3. Assume that the resource adapter is deployed prior to the MDB. (The MDB could be deployed before the resource adapter is deployed; in that case, the deployed MDB polls until the resource adapter is deployed.) When the resource adapter is deployed, the ResourceAdapter bean is bound into JNDI using the name specified.

  4. The MDB is deployed, and the MDB container invokes an application server-specific API that looks up the resource adapter by its JNDI name and invokes the specification-mandated endpointActivation(MessageEndpointFactory, ActivationSpec) method on the resource adapter.

  5. The MDB container provides the resource adapter with a configured ActivationSpec (containing configuration information) and a factory for the creation of message endpoint instances.

  6. The resource adapter saves this information for later use in message delivery. The resource adapter thereby knows what message listener interface the MDB implements. This information is important for determining what kind of messages to deliver to the MDB.

Dispatching a Message

When a message arrives from the EIS to the resource adapter, the resource adapter determines where to dispatch it. The following is a possible sequence of events:

  1. A message arrives from the EIS to the resource adapter.

  2. The resource adapter examines the message and determines its type by looking it up in an internal table. The resource adapter determines the message type corresponds to a particular pair (MessageEndpointFactory, ActivationSpec).

  3. The resource adapter determines the message should be dispatched to an MDB.

  4. Using the MessageEndpointFactory for that type of message endpoint (one to be dispatched to an MDB), the resource adapter creates an MDB instance by invoking createEndpoint() on the factory.

  5. The resource adapter then invokes the message listener method on the MDB instance, passing any required information (such as the body of the incoming message) to the MDB.

  6. If the message listener does not return a value, the message dispatching process is complete.

  7. If the message listener returns a value, the resource adapter determines how to handle that value. This may or may not result in further communication with the EIS, depending upon the contract with the EIS.

Activation Specifications

A resource adapter is configured with a mapping of message types and activation specifications. The activation specification is a JavaBean that implements javax.resource.spi.ActivationSpec. The resource adapter has an ActivationSpec class for each supported message type. The mapping of message types and activation specifications is configured in the ra.xml deployment descriptor, as described in Configuring Inbound Connections, For more information about ActivationSpecs, see Chapter 13, "Message Inflow," in JSR 322: Java EE Connector Architecture 1.6.

Administered Objects

As described in section 13.4.2.3 of JSR 322: Java EE Connector Architecture 1.6, a resource adapter may provide the Java class name and the interface type of an optional set of JavaBean classes representing administered objects that are specific to a messaging style or message provider. You configure administered objects in the admin-objects elements of the ra.xml and weblogic-ra.xml deployment descriptor files. As with outbound connections and other WebLogic resource adapter configuration elements, you can define administered objects at three configuration scope levels:

  • Global - Specify parameters that apply to all administered objects in the resource adapter using the default-properties element. See default-properties in Table A-15, "admin-object-group"

  • Group - Specify parameters that apply to all administered objects belonging to a particular administered object group specified in the ra.xml deployment descriptor using the admin-object-group element. The properties specified in a group override any parameters specified at the global level. See admin-object-group.

    The admin-object-interface element (a subelement of the admin-object-group element) serves as a required unique element (a key) to each admin-object-group. There must be a one-to-one relationship between the admin-object-interface element in weblogic-ra.xml and the admin-object-interface element in ra.xml.

  • Instance - Under each admin object group, you can specify administered object instances using the admin-object-instance element of the weblogic-ra.xml deployment descriptor. These correspond to the individual administered objects for the resource adapter. You can use the admin-object-properties subelement to specify properties at the instance level too; properties specified at the instance level override those provided at the group and global levels. See admin-object-instance.

Transactional Inflow

This section discusses how transactions flow into WebLogic Server from an EIS and a resource adapter. A transaction inflow contract allows the resource adapter to handle transaction completion and crash recovery calls initiated by an EIS. It also ensures that ACID properties of the imported transaction are preserved. For more information on transaction inflow, see Chapter 15, "Transaction Inflow," in JSR 322: Java EE Connector Architecture 1.6.

When an EIS passes a message through a resource adapter to the application server, it may pass a transactional context under which messages are delivered or work is performed. The inbound transaction will be controlled by a transaction manager external to the resource adapter and application server. See Message Inflow to Message Endpoints (Message-driven Beans).

A resource adapter may act as a bridge between the EIS and the application server for transactional control. That is, the resource adapter receives messages that it interprets as XA callbacks for participating in a transaction with a external Transaction Manager.

WebLogic Server can function as an XA resource to a external Transaction Manager through its interposed Transaction Manager. The WebLogic Server Transaction Manager maps external transaction IDs to WebLogic Server-specific transaction IDs for such transactions.

The WebLogic Server Transaction Manager is subordinate to the external Transaction Manager, which means that the external Transaction Manager ultimately determines whether the transaction succeeds or is rolled back. See "Participating in Transactions Managed by a Third-Party Transaction Manager" in Oracle Fusion Middleware Programming JTA for Oracle WebLogic Server. As part of the Java EE Connector Architecture 1.6, the ability for a resource adapter to participate in such a transaction is now exposed through a Java EE standard API.

The following process explains how a resource adapter would participate in a external transaction. For more information, see section 15.4, "Transaction Inflow Model," in JSR 322: Java EE Connector Architecture 1.6.

  1. The resource adapter receives an inbound message with the transaction context that arrived along with the incoming message.

  2. The resource adapter represents the transaction context using the javax.transaction.xa.Xid instance.

  3. The resource adapter creates a new Work instance to process the incoming message and deliver it to a message endpoint, and also creates an instance of an ExecutionContext (javax.resource.spi.work.ExecutionContext), setting the Xid it created and also setting a transaction timeout value. Version 1.6 of the Connector Architecture defines a standard class, TransactionContext, which resource adapters may use instead of the ExecutionContext for propagating the transaction context from the EIS to the application server.

  4. The resource adapter submits the Work object and the TransactionContext (or ExecutionContext) to the Work Manager for processing. At this point, the Work Manager performs the necessary work to enlist the transaction and start it with the WebLogic Server Transaction Manager.

    To use a TransactionContext, the Work class must:

    1. Implement the javax.resource.spi.work.WorkContextProvider interface.

    2. Create and return a TransactionContext instance in the getWorkContexts() method.

    Note:

    If the resource adapter uses a TransactionContext, the adapter must not use an ExecutionContext.

  5. Subsequent XA calls from the external Transaction Manager are sent through the resource adapter and communicated to the WebLogic Server Transaction Manager. In this way, the resource adapter acts as a bridge for the XA calls between the external Transaction Manager and the WebLogic Server Transaction Manager, which is acting as a resource manager.

Using the Transactional Inflow Model for Locally Managed Transactions

When the resource adapter receives requests from application components running in the same server instance as the resource adapter that need to be delivered to an MDB as part of the same transaction as the resource adapter request, the transaction ID must be obtained from the transaction on the current thread and placed in a TransactionContext (or ExecutionContext).

In this case, WebLogic Server does not use the Interposed Transaction Manager but simply passes the transaction on to the Work Thread used for message delivery to the MDB.

Configuring and Managing Long-Running Work

As mentioned in Long-Running Work Support, WebLogic Server supports the use of HintsContext.LONGRUNNING_HINT. If set to true in a resource adapter Work class, LONGRUNNING_HINT causes a Work instance to be established as a long-running work item that WebLogic Server schedules in a separate daemon thread, not in a Work Thread. LONGRUNNING_HINT performs the same function as the WebLogic Server extension annotation @LongRunning.

WebLogic Server extends Connector Architecture 1.6 by providing the ConnectorWorkManagerRuntimeMBean, which contains attributes for configuring and monitoring long-running Work instances. These attributes, described in the following sections, are also exposed in the WebLogic Server Administration Console.

For more information about the @LongRunning extension annotation, see LongRunning in Oracle Fusion Middleware Oracle WebLogic Server 10.3.1 API Reference.

Setting the Maximum Number of Concurrent Long-Running Work Instances

Oracle recommends that you minimize the number of long-running Work instances executing concurrently because each long running work runs in its own daemon thread. Having too many concurrent long-running Work instances can exhaust the thread resources in WebLogic Server and cause a negative impact on server performance and stability. WebLogic Server may introduce restrictions on maximum concurrent long running works allowed in a future release.

You can use the WebLogic Server Administration Console to set the maximum allowed number of concurrent Work instance requests as follows:

  1. Select the resource adapter in the Summary of Deployments > Control page.

  2. Select Configuration > Workload.

  3. Enter a new value in Maximum Number of Concurrent Long Running Requests, if desired, and click Save.

    If you save a new value, the Save Deployment Plan Assistant is displayed, which prompts you to select or enter the path of a deployment plan file. For more information about working with deployment plans, see "Understanding WebLogic Server Deployment" in Deploying Applications to Oracle WebLogic Server.

Note the following:

  • You can also view the maximum number of concurrent Work instance requests allowed from the Resource Adapter: Monitoring: Workload page in the Administration Console, as described in Monitoring Long-Running Work.

  • As an alternative to using the Administration Console, you can use the max-concurrent-long-running-requests element in the weblogic-ra.xml file to set the maximum allowed number of concurrent Work instance requests. For information, see connector-work-manager.

Monitoring Long-Running Work

The ConnectorWorkManagerRuntimeMBean exposes long-running run-time information about the resource adapter's specific Work Manager in the following MBean attributes:

To view information about the currently active or completed long-running Work instance requests using the WebLogic Server Administration Console:

  1. Select the resource adapter in the Summary of Deployments > Control page.

  2. Select Monitoring > Workload.

    The following information about long-running Work instance requests is available from the Long Running Work Managers table:

    The column labeled . . . . . . identifies the following

    Active Requests

    The number of currently active long-running Work instance requests.

    Completed Requests

    The number of completed long-running Work instance requests.

    Max Concurrent Requests Allowed

    The maximum number of concurrent Work instance requests allowed.