Lot of JMS/SIB communication beetwen servers in cluster
Noticed some Chordiant activity on JMS and SIB components on WebSphere Application Server. In our solution we suspect that there is no need for any data to be synchronized by Chordiant between servers in cluster (at the moment we have 2 nodes with 3 app servers on each). Users should be sticked to one server and their http sessions aren't replicated.
Is it possible to disable any such communication for Chordiant? Its been noticed that it takes noticeable amount of CPU to execute onMessage metods related to ex. UserProfile_Cache_Topic
During diagnostic of performance problem we are changing number of concurrently running application servers. Can it cause any problems or overhead if number of all servers in config file is bigger than real number of servers/instances?
We are using Chordiant 6.6 on WebSphere Application Server 7 on AIX.
Steps to Reproduce
Log in to Chordiant
The root cause of the issue was number of JMS messages being passed between servers in Cluster
Given the uniques cluster requirement at customers end, following suggestion and a tuning patch to mitigate the problem has been provided:
1. Setting RegistryConnector.xml \ SYNCACHE parameter to false
Please note: SYNCCACHE set to true synchronizes UserProfileCache across cluster, at the time of failover the user activity are not affected.
If you set this to false, UserProfileCache is not updated across cluster, so if you add roles or permission to user from a different machine, its very likely that the cache in other jvm’s won’t be updated and changes will not reflect.
However, subsequent re-start of cluster will pick up new values from DB and then changes will be reflected. This is a trade-off, if you set SYNCCACHE to false.
2. Applying tuning patch HotFix-912 (modifications of SessionTopicListener java class).
3. Setting "Maximum concurrent MDB invocations per endpoint" parameter value to 1 in WAS admin console.
0% found this useful