I've been trying hard to configure a 3 node Infinispan cluster, which uses Infinispan and Hibernate Search 3.4.1. I configured Infinispan to use a shared JDBC backed CacheLoader, in a MySQL database.
Sincerely, after a couple of weeks I am not able to get a stable configuration, and I am having different exceptions and frequent lucene index corruption. Please see my comments in:
At first I tried to configure Infinispan 5.0.1.FINAL, with JGroups 2.12. But "Read past EOF: Chunk value could not be found for key" exceptions arised frequently when I restarted the nodes, bringing my lucene indexes unusable, and requiring a complete reindex of all my entities from scratch.
Then I found in forums that Infinispan 5.0.1.FINAL is not working well with HSearch 3.4.1, so I downgraded to 4.2.1, as suggested here:
Now I have new "Read past EOF exceptions" restarting my nodes. (Read past EOF at org.infinispan.lucene.SingleChunkIndexInput.readByte).
So, I am asking if someone is using this technology in a production environment, because based on this experience I am considering switching to a conventional HSearch JMS Cluster. Maybe I have a misconfiguration issue, but I did tried several configurations with the same unexpected results.
My current Infinispan and JGroups configuration is based on the configuration posted by pbratton in the previous forum link.
My current environment:
- JBoss 4.2 AS
- Infinispan 4.2.1.FINAL
- JGroups 2.11.1.FINAL
- Hibernate Search 3.4.1.Final
- Lucene-core 3.1.0
- MySQL 5.0 as JDBC CacheStore/loader
Any help would be really appreaciated.