[jira] [Comment Edited] (AMQ-5694) ActiveMQTempDestination.delete() can block forever

["Matt Pavlovich (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/AMQ-5694?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17294682#comment-17294682 ] 

Matt Pavlovich edited comment on AMQ-5694 at 3/3/21, 5:32 PM:
--------------------------------------------------------------

Note: This ticket is being marked for closure due to age and inactivity.

Please provide an update if this is still an issue against 5.16.1 or newer. There has been quite a bit of work on the JDBC backend in the 5.16.x releases.

Information need to help identify the root cause:
1. ActiveMQ version
2. ActiveMQ configuration file
3. Database type and vendoer
4. JDBC driver jar and version
5. JDBC connection url 


was (Author: mattrpav):
Please provide an update if this is still an issue. This ticket is being marked for closure due to age and inactivity.

Information need to help identify the root cause:
1. ActiveMQ version
2. ActiveMQ configuration file
3. Database type and vendoer
4. JDBC driver jar and version
5. JDBC connection url 

> ActiveMQTempDestination.delete() can block forever
> --------------------------------------------------
>
>                 Key: AMQ-5694
>                 URL: https://issues.apache.org/jira/browse/AMQ-5694
>             Project: ActiveMQ
>          Issue Type: Bug
>            Reporter: Endre Stølsvik
>            Assignee: Matt Pavlovich
>            Priority: Major
>              Labels: close-pending
>         Attachments: thread hangs.png
>
>
> As described in AMQ-5681, we have a setup where we every 10 seconds queries the broker over JMS for a StatisticsMessage.
> This bug concerns a .. somewhat related .. problem: In that code path, we delete the temporary reply-to queue after we've read the data.
> We've now several times ended up with a peculiar situation where the thread seems to have died. Today I grabbed a JMX console and had a look, and the thread is not dead, it is just locked on the delete call. (see screenshot)
> We have another problem with a shared JDBC "cluster" (the "single master, hot standbys" setup), where the nodes loose master and effectively goes down. What I believe happens, is that the thread sends the delete message, and then goes into a blocking wait for the reply, which never appears, probably because the broker that was master and should have sent it, is now dead.
> The delete code path should probably have had some sane timeout, and just have raised some JMSException - or something.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)