You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@qpid.apache.org by "Rob Godfrey (JIRA)" <ji...@apache.org> on 2014/06/10 11:27:02 UTC
[jira] [Updated] (QPID-5809) [Java broker] Investigate 1.0
connection clean up process
[ https://issues.apache.org/jira/browse/QPID-5809?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Rob Godfrey updated QPID-5809:
------------------------------
Description:
>From comment by @gemmellr on QPID-5771:
The issue is that the severed connection is not removed from the ConnectionRegistry when it attempts to clean itself up, because its cleanup process itself throws an exception after trying to use the connection again to close the sessions. The ConnectionRegistry then tries to close the connection, and gets stuck in a loop because the close attempt isn't actually doing anything (the Connection is already marked closed) that could cause it to be removed from the registry, so the shutdown hook never completes and it will sit there as observed.
The 1.0 session etc objects in/on the connection get cleaned up within specific 1.0 connection closed() cleanup method I found was failing to unregister the connection (its the only place that does that and it was failing to do it) ... in the case I saw, it was trying to send something over the connection object that was closing itself.
was:... if the connection from the attached test is killed (rather than closing cleanly).
> [Java broker] Investigate 1.0 connection clean up process
> ---------------------------------------------------------
>
> Key: QPID-5809
> URL: https://issues.apache.org/jira/browse/QPID-5809
> Project: Qpid
> Issue Type: Bug
> Components: Java Broker
> Affects Versions: 0.28
> Reporter: Rob Godfrey
> Assignee: Rob Godfrey
> Fix For: 0.29
>
> Attachments: Simple.java
>
>
> From comment by @gemmellr on QPID-5771:
> The issue is that the severed connection is not removed from the ConnectionRegistry when it attempts to clean itself up, because its cleanup process itself throws an exception after trying to use the connection again to close the sessions. The ConnectionRegistry then tries to close the connection, and gets stuck in a loop because the close attempt isn't actually doing anything (the Connection is already marked closed) that could cause it to be removed from the registry, so the shutdown hook never completes and it will sit there as observed.
> The 1.0 session etc objects in/on the connection get cleaned up within specific 1.0 connection closed() cleanup method I found was failing to unregister the connection (its the only place that does that and it was failing to do it) ... in the case I saw, it was trying to send something over the connection object that was closing itself.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@qpid.apache.org
For additional commands, e-mail: dev-help@qpid.apache.org