Since 6.9.1 P02 and as well in 7.0.2 P01 HF01 (both on Websphere) we get the following error message frequently in our AveksaServer.log:
The JNDI connection cache already contains key - 'com.ibm.ws.rsadapter.jdbc.WSJdbcConnection@1298de89312008329'. This could be a problem.
Does anyone know what that means?
Do we need to fix it?
Those workflows are currently in production with 6.9.1 P02 HF07.
This is a clean 7.0.2 P01 HF01 installation with less than 10 requests in the system.
We have not seen any error in the log yet that would show us a problematic node or anything like that.
Workflows are stalling and pausing randomly.
I agree that those possible bad workflows need to be identified and resolved but with the given information by the product it is nearly impossible for us.
Performance related issues are hard to troubleshoot by definition, as there typically is no error message as there is no logical fault. Advanced techniques need to be used to identify which workflows might be inefficient or looping. This is best done through a support case which I see you have open for this issue.
I would expect this is an artifact of your other issues related to transaction rollback failures. This suggest the system is having failures and then been restarted, and having problems recovering the failed transactions. If you clean up all of the other errors this should go away.
It sounds like this environment has performance issues. Likely there are some bad workflows that are causing a very high failure load in workpoint. These all need to be identified and resolved.