SourceForge.net
2005-05-03 17:37:04 UTC
Bugs item #1194602, was opened at 2005-05-03 19:37
Message generated for change (Tracker Item Submitted) made by Item Submitter
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=474136&aid=1194602&group_id=54559
Category: message engine
Group: v0.7.6.1
Status: Open
Resolution: None
Priority: 5
Submitted By: tishod (tishod)
Assigned to: Jim Alateras (jalateras)
Summary: Memory leak when browsing a queue
Initial Comment:
The leak happens in the server when performing a lot of
browsing on a point-to-point queue.
The queue has several consumers that listen
synchronously. A producer sends at high speed messages
via the RMI protocol. The messages are configured to
expire after 1 minute so that the server does not
explode with OutOfMemory exception. A browsing client
browses the queue at regular intervals (10 seconds) and
when the queue depth reaches some minimum (100
messaages) it inserts several new messages (900) in the
same way as the producer.
The producers are faster than the consumers but are
slower than the expiration of the inserted messages.
I.e. messages are inserted at high speed but they
expire after 1 minute or are consumed.
After 10-30 minutes (depending on the JVM heap size)
the server explodes with OME. I am watching the queue
with the admin tool and I see that at the moment when
the server explodes with OEM the queue is far from full
- i.e. the explosion is not caused by overfilled queue.
When I do the same test but without starting the
browing client the server does not explode.
As workaround for this bug I started to use the
administration interface to ask for the depth of the
queue. But there are two problems with it:
1. I can not know which messages are still in the queue
which is a problem in my case.
2. The administration interface is not a JMS standard.
----------------------------------------------------------------------
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=474136&aid=1194602&group_id=54559
Message generated for change (Tracker Item Submitted) made by Item Submitter
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=474136&aid=1194602&group_id=54559
Category: message engine
Group: v0.7.6.1
Status: Open
Resolution: None
Priority: 5
Submitted By: tishod (tishod)
Assigned to: Jim Alateras (jalateras)
Summary: Memory leak when browsing a queue
Initial Comment:
The leak happens in the server when performing a lot of
browsing on a point-to-point queue.
The queue has several consumers that listen
synchronously. A producer sends at high speed messages
via the RMI protocol. The messages are configured to
expire after 1 minute so that the server does not
explode with OutOfMemory exception. A browsing client
browses the queue at regular intervals (10 seconds) and
when the queue depth reaches some minimum (100
messaages) it inserts several new messages (900) in the
same way as the producer.
The producers are faster than the consumers but are
slower than the expiration of the inserted messages.
I.e. messages are inserted at high speed but they
expire after 1 minute or are consumed.
After 10-30 minutes (depending on the JVM heap size)
the server explodes with OME. I am watching the queue
with the admin tool and I see that at the moment when
the server explodes with OEM the queue is far from full
- i.e. the explosion is not caused by overfilled queue.
When I do the same test but without starting the
browing client the server does not explode.
As workaround for this bug I started to use the
administration interface to ask for the depth of the
queue. But there are two problems with it:
1. I can not know which messages are still in the queue
which is a problem in my case.
2. The administration interface is not a JMS standard.
----------------------------------------------------------------------
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=474136&aid=1194602&group_id=54559