Tips for troubleshooting your WS-ReliableMessaging configuration.
To help you identify and resolve problems with WS-ReliableMessaging, you can use the TCP/IP monitor to view the messages that are flowing between your client applications and reliable messaging enabled Web services. You can also use the WebSphere® Application Server trace and logging facilities as described in Tracing and logging configuration.
org.apache.sandesha2*=all=enabled:com.ibm.ws.websvcs.rm*=all=enabled:org.apache.axis2*=all=enabled:com.ibm.ws.sib.wsrm*=all=enabled
org.apache.sandesha2*=all=enabled:com.ibm.ws.websvcs.rm*=all=enabled:org.apache.axis2*=all=enabled
A list of the main known restrictions that apply when using WS-ReliableMessaging is provided in WS-ReliableMessaging known restrictions.
WebSphere Application Server system messages are logged from a variety of sources, including application server components and applications. Messages logged by application server components and associated IBM® products start with a unique message identifier that indicates the component or application that issued the message. The prefix for the WS-ReliableMessaging component is CWSKA.
The Troubleshooter reference: Messages topic contains information about all WebSphere Application Server messages, indexed by message prefix. For each message there is an explanation of the problem, and details of any action that you can take to resolve the problem.
If a sequence is reallocated, the original and new sequences are both visible. Ignore the multiple entries.
This is because, although reliable messaging only binds to one messaging engine in a cluster, the runtime panel is calculating and displaying the sequence information once for every cluster member. Ignore the duplicate entries. Note that the slight differences in the statistics being displayed for each duplicate entry is due to the entries being created sequentially, while polling for messages continues.
If you are migrating from WebSphere Application Server Version 6.1, and you are using Version 6.1.0.9 or 6.1.0.11 of the Feature Pack for Web Services, and your configuration includes WS-ReliableMessaging configured for the managed persistent quality of service, you need to remove all persisted messages before you migrate.
When you use reliable messaging with a managed quality of service, you might see the following exception message when your application server starts:
CWSIT0019E: No suitable messaging engine is available on bus yourBus that matched the specified connection properties
In a network deployment environment, this can occur because the messaging engine is on an application server or cluster member that has started later than the server that hosts your reliable messaging application. In this case you need do nothing but wait; reliable messaging will keep trying to connect until the messaging engine becomes available.
After a sequence has been established, WS-ReliableMessaging provides retransmission of messages to a service. However if the sequence is not established then the messages are not transmitted to the service and a message similar to the following example is displayed:
org.apache.axis2.AxisFault: The Create Sequence request has been refused by the RM Destination
The initial createSequence message has been refused. This is propagated back, and causes the client to fail. For information about CreateSequence and CreateSequenceRefused, see the WS-ReliableMessaging: supported specifications and standards.
You might also see a subsequent message to help explain why the request has been refused. For example:
Caused by: javax.xml.ws.soap.SOAPFaultException: com.ibm.ws.sib.wsrm.exceptions.WSRMRuntimeException: CWSJZ0202I: A messaging engine connection is unavailable for bus myBus.
If you get an exception such as the following exception, then the sequence is established but cannot be used:
javax.xml.ws.WebServiceException: org.apache.axis2.AxisFault: The value of wsrm:Identifier is not a known Sequence identifier.
The most common reason is that you are working in a clustered environment but your server-side policy set specifies the unmanaged non-persistent quality of service. For example: The WS-I RSP default policy set specifies the unmanaged non-persistent quality of service. To use reliable asynchronous messaging in a clustered environment, you must use a managed quality of service to enable the cluster members to correlate reliable messaging state. To do this, either use the WS-I RSP ND default policy set, or modify your custom policy set so that the WS-ReliableMessaging policy specifies a managed quality of service, and an associated binding to a service integration bus and messaging engine. For information about how to do this, see Configuring a WS-ReliableMessaging policy set by using the administrative console and Attaching and binding a WS-ReliableMessaging policy set to a Web service application by using the administrative console.
If your policy set specifies a managed quality of service, but you have not specified a binding to a messaging engine to support that quality of service, you get the following exception message:
CWSKA0102E: The managed Web services reliable messaging storage manager could not be initialized because the policy set binding was incomplete or invalid.
Perhaps you have attached a managed policy set to your application, and used the default bindings (which do not support the managed qualities of service). You must create a new binding for your application that specifies a service integration bus and messaging engine to support the managed qualities of service. To do this, see Attaching and binding a WS-ReliableMessaging policy set to a Web service application by using the administrative console.
Clustering offers maximum protection against servers becoming unavailable. It provides highly available service endpoints, and (through the service integration bus) high availability of the reliable messaging layer.
For more information about configuring high availability for Web services and messaging engines, see Balancing workloads and Adding a messaging engine to a cluster.
When many applications use the same messaging engine, it can impact performance and in some cases lead to timeout errors.
For more information about configuring high availability for Web services and messaging engines, see Balancing workloads and Adding a messaging engine to a cluster.
When the reliable messaging layer receives a request message, it sends an acknowledgement then delivers the message to the target service. There is a marginal possibility that the server hosting the reliable messaging layer might become unavailable after the request message has been acknowledged and before it has been delivered. In this case, the message is only recovered if you are using in-order delivery as well as managed persistent quality of service. To specify in-order delivery, select the WS-ReliableMessaging policy option to "Deliver messages in the order that they were sent" as described in Configuring the WS-ReliableMessaging policy.
CWWSS7215E: Cannot get valid security context token from the cache.
To ensure that the scoping security context token does not expire before WS-ReliableMessaging can recover and resend its messages, complete the following task: Configuring WS-SecureConversation to work with WS-ReliableMessaging.
There is a performance overhead in using in-order delivery, because messages are held in a queue until they can be delivered in order. In some situations, messages can build up in the queue until an out-of-memory error occurs. If a client is sending messages to the server, but not receiving responses, you can configure the client to stop sending messages until the server recovers, preventing the build up of messages on the server. You can also tune your system to increase the heap size of the JVM, which can help to avoid out-of-memory errors.
However, where the highest level of reliability is required, you should always specify in-order delivery in conjunction with the managed persistent quality of service.
The performance overhead
is greatly reduced; you should not have to tune your system to reduce
its effect, although you might want to tune your system for other
purposes. You are unlikely to see out-of-memory errors resulting from
the use of in-order delivery.