I have a Bayeux implementation as specified by IBM. In a stand-alone environment it works fine, but when we put it into production (clusterized), sessions are getting lost after a while. We are using SIBUS as the bus implementation and the default messaging provider. In SIBUS we are using a database based data store. Our web server is WebSEAL.
We are sure this is not a replication issue since:
- Sticky sessions are enabled
- After start-up, every participant is able to see each other, which means that it doesn't matter where you signed-in, session is being replicated by both servers.
The main symptom is that after a few minutes of working OK, the long-polled request hangs until a new request originated at the client side happens. There are no exceptions shown anywhere.
I still cannot tell if my problem is with the SIBUS configuration, or if it would have something to do with the WebSEAL configuration.
Any help will be really appreciated. IBM documentation is not the best and we have been testing every piece of advice found in several documents with no luck.
EDIT : After some testing, I have found that the behavior happens also on non-clusterized environments. I'm also seeing an exception with the code CWPSB1314W, which gives me nothing.