Bela, thank you for answer! It seem to me, that there could be at least two sources of such problems.
1. JVM crashes within the replication stage. We kill it under high load (a lot of replications), but sometimes we should do this 5-7 times to have such effect. This is right what JBCache-10 you mentioned is about. 2. This problem can be connected to transaction manager. We use JOTM. As I can understand from it's sources - if transaction is timed out, it's removed from thread local, and when rollback is called, IllegalStateException arises. Thus, transaction is not rolled back and nodes remain blocked. For now we are investigating this possibility. We can avoid second problem by playing with timeouts, but the first still remain. It would be VERY nice if TreeCache would be able to release all locks from crashed members, since doing this manualy is quite heavy procedure. I'll try TreeCache.removeAllLocks(Fqn) within a couple of hours. Regards, Eugene aka Skipy View the original post : http://locahost:8080/index.html?module=bb&op=viewtopic&p=3892399#3892399 Reply to the post : http://locahost:8080/index.html?module=bb&op=posting&mode=reply&p=3892399 ------------------------------------------------------- SF.Net email is Sponsored by the Better Software Conference & EXPO September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf _______________________________________________ JBoss-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/jboss-user
