activemq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Claus Ibsen (Updated) (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (AMQ-3588) PooledSession - Can cause memory leak in case many consumers is created by the session
Date Thu, 10 Nov 2011 17:10:53 GMT

     [ https://issues.apache.org/jira/browse/AMQ-3588?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Claus Ibsen updated AMQ-3588:
-----------------------------

    Attachment: AMQ-3588.patch

The patch
                
> PooledSession - Can cause memory leak in case many consumers is created by the session
> --------------------------------------------------------------------------------------
>
>                 Key: AMQ-3588
>                 URL: https://issues.apache.org/jira/browse/AMQ-3588
>             Project: ActiveMQ
>          Issue Type: Bug
>          Components: activemq-camel
>            Reporter: Claus Ibsen
>            Assignee: Claus Ibsen
>            Priority: Critical
>              Labels: activemq-pool, camel,
>             Fix For: 5.6.0
>
>         Attachments: AMQ-3588.patch, after.png, before.png
>
>
> See (CAMEL-4657
> In the activemq-pool we have a memory leak when a PooledSession is used to create many
consumers. As each consumer is added to an internal list, which in case the session is closed,
the list is used to ensure all previously created consumers gets closed as well.
> However if the session is used to create many consumers, then that internal list just
keep growing, and causing OOME. See attached screenshots of a before vs after with the bug
fix.
> So to remedy this situation, we should ensure any closed consumer gets removed from that
internal list. As its already closed, and therefore the session do not need to keep track
of it anymore.
> This bug have existed ever since AMQ-615.
> The impact of this bug can be seen in CAMEL-4657, which essentially makes using Camel
for request/reply over JMS with AMQ and with named replyTo queues (eg not temporary) will
cause this problem. The problem manifests due Spring DMLC, which uses a MessageSelector to
only pickup expected reply messages. And because the MessageSelector is created once when
the consumer is created. We need to create a new Consumer, every time a new message is doing
request/reply, to ensure the Message Selector gets updated. So in that situation many consumers
get created, which leads to OOME as the list in PooledSession keeps growing.
> It would be great if this bug fix can be backported to the 5.5 branch as well.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message