Bugs item #456257, was opened at 2001-08-28 12:45 You can respond by visiting: http://sourceforge.net/tracker/?func=detail&atid=376685&aid=456257&group_id=22866
Category: JBossServer Group: v3.0 Rabbit Hole Status: Open Resolution: None Priority: 7 Submitted By: Bill Burke (patriot1burke) Assigned to: Nobody/Anonymous (nobody) Summary: entity may not be released on TX timeout Initial Comment: > JBoss uses pessimistic locking. So as an entity bean is accessed within a > transaction it is pessimistically locked. > > What if the thread that is the transaction is tied up blocking forever on > I/O? Yes, the TM will mark the transaction as rolled back, but the entity > beans locked within the transaction will not be released because > the thread > will never wakeup and realize it must do a rollback. Am I wrong here? ---------------------------------------------------------------------- >Comment By: Jason Dillon (user57) Date: 2002-04-05 23:31 Message: Logged In: YES user_id=15045 Ok, fixed yet? ---------------------------------------------------------------------- Comment By: David Jencks (d_jencks) Date: 2002-02-05 10:09 Message: Logged In: YES user_id=60525 Just saw this... I assume "resources the bean has open" == resource adapter connections. I am working on having the container keep track of these. Can someone point out where the timeout notification comes from? Thanks david jencks ---------------------------------------------------------------------- Comment By: Ole Husgaard (sparre) Date: 2001-10-18 08:35 Message: Logged In: YES user_id=175257 This is correct. The real problem here is a thread that is hung in the bean code. Probably, the container should somehow detect this and take appropriate measures, like killing the thread, closing all resources the bean instance has open and discarding the bean instance as if it threw a RuntimeException. Also just marking the transaction for rollback only is IMHO not enough. As it is still active, XA resources may heuristically complete the transaction, and their decision to commit or rollback is beyond our control. If a resource decides to do a heuristic commit, we may end up with a mixed heuristic that breaks the atomicy property of transactions. A better approach would be to have the transaction do a heuristic rollback on timeout, and make the container kill any threads executing in bean instances on that transaction and close all resources the bean has open and discard these bean instances. But I think that the first TODO here is making the container keep track of the resources that bean instances have open. ---------------------------------------------------------------------- Comment By: Bill Burke (patriot1burke) Date: 2001-10-01 07:32 Message: Logged In: YES user_id=176497 This has definately NOT been fixed. ---------------------------------------------------------------------- Comment By: Jason Dillon (user57) Date: 2001-09-28 15:50 Message: Logged In: YES user_id=15045 Do you know if this has been fixed? If so can we close this & mark it as fixed? I kinda remember some talk about this being fixed, but I can remember. ---------------------------------------------------------------------- You can respond by visiting: http://sourceforge.net/tracker/?func=detail&atid=376685&aid=456257&group_id=22866 _______________________________________________ Jboss-development mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-development
