Rajika has written a very comprehensive article about setting up WSO2 ESB with JMS in a fail-over scenario.
The concept of fail-over endpoints is integral to WSO2 ESB. The user can define a fail-over endpoint in the ESB which in turns consists of two or more child endpoints. ESB will select one of the child endpoints to be the primary endpoint and will keep forwarding all messages to this primary endpoint. If the ESB detects some error while communicating with the primary endpoint, it will automatically fail-over to one of the available backup endpoints. User can also configure the endpoint error codes upon which to trigger a fail-over, endpoint suspend durations and timeout durations. However this is not what Rajika is talking about in his article. His article is about setting up a JMS proxy service with fail-over support in the ESB .
For an example imagine a situation where you have setup a proxy service to receive messages from a JMS queue. The ESB proxy service polls messages out of the queue and forwards them to a backend application for further processing. Now what if the ESB instance which hosts the above mentioned proxy service happens to fail (say due to a hardware failure). Actually in this case there's both good news and bad news. Good news is that no messages will get lost due to the persistent behavior of JMS messaging. If some messages were in flight within the ESB when the failure occurred those messages might get lost. But even that can be avoided by using something like JTA. The bad news however is that the backend application will not receive further requests to be processed. So as far as your overall system is concerned, your applications and services will stay idle while messages keep piling up on the JMS queue.
Rajika discusses a possible architecture to avoid such system downtime by having a backup proxy service (ideally deployed in a separate ESB instance) which listens on the same JMS queue. The backup proxy will be disabled by default to prevent contention between the two proxy services. His architecture also utilizes a scheduled task deployed on the backup ESB which sends periodic heartbeats to the other ESB node. If the backup ESB detects that the primary ESB instance is not responding, it will activate the backup proxy service which will start polling for JMS messages and resume operations in the system.
This is actually fairly simple to implement in WSO2 ESB and can be even further customized to suit the exact enterprise requirements. All the source code and relevant configurations are provided in the article itself.