[openjms-developer] [ openjms-Bugs-768543 ] Deadlock btw LeaseManager and DestinationManager.collectGarb

[SourceForge.net]

Bugs item #768543, was opened at 2003-07-10 01:41
Message generated for change (Comment added) made by tanderson
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=474136&aid=768543&group_id=54559

Category: server
Group: v0.7.5

Status: Closed

Resolution: Fixed
Priority: 5
Submitted By: David Magro Martin (dmagrom)
Assigned to: Tim Anderson (tanderson)
Summary: Deadlock btw LeaseManager and DestinationManager.collectGarb

Initial Comment:
We think we have discovered a deadlock problem....

We have 5 clients connected to our openjms server. At
the moment we are not using any MessagerListeners in
the consumer threads, but rather doing continous
polling (receiveNoWait) with a 5ms interval.

Since we have no listeners and do no receive, there are
intervals with no consumer threads attached to the
corresponding destination cache. So, every second
(default setting) the cache gets garbaged-collected,
but at the same time there are expired messages in the
cache and the
lease manager tries to remove them. This produces
a deadlock as can be seen in the stack trace (attached
file).

All our messages are persistent (mysql backend) and
some of them have a TTL.

Thank you very much.

----------------------------------------------------------------------

Comment By: Tim Anderson (tanderson)

Date: 2003-08-12 12:19

Message:
Logged In: YES
user_id=557161

Fix will be available in the 0.7.6 release.

----------------------------------------------------------------------

Comment By: Tim Anderson (tanderson)
Date: 2003-07-22 18:03

Message:
Logged In: YES
user_id=557161

Tentatively fixed in CVS. I couldn't reproduce the original
problem, but I've reduced the level of synchronization in
LeaseManagerHelper.

Could you check it out and see if it fixes the problem?

Thanks,

Tim

----------------------------------------------------------------------

Comment By: David Magro Martin (dmagrom)
Date: 2003-07-15 19:10

Message:
Logged In: YES
user_id=82773

Sorry, but we cannot reproduce it either. It just happens
very seldom, about once in 10 days but not regularly.
The server is running non-stop with 6 queues (no topics)
under more or less heavy load (10 msgs per second, some of
them around 40Kbytes )

Thank you.

----------------------------------------------------------------------

Comment By: Tim Anderson (tanderson)
Date: 2003-07-14 22:40

Message:
Logged In: YES
user_id=557161

I'm having trouble reproducing this. I can make changes
based on the stack trace but can't guarantee they will fix the
problem. Can you provide a reproducible test case?

----------------------------------------------------------------------

You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=474136&aid=768543&group_id=54559