[jboss-jira] [JBoss JIRA] (JBMESSAGING-1945) Clustered queues get deleted from the jbm_postoffice table when it should not

[Yong Hao Gao (JIRA)]

    [ https://issues.jboss.org/browse/JBMESSAGING-1945?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12782081#comment-12782081 ] 

Yong Hao Gao commented on JBMESSAGING-1945:
-------------------------------------------

Here is my observation:

First when you do 'ifdown', the remote node will be failed over. As part of the failover the entries for the remote
queues will be removed.

Then when you do 'ifup', the remote node won't be re-joining the cluster as it has been failed over already, in which
case the only way to re-join is restart the remote node.

After that although both node are up, the cluster's state is actually not correct. For example the removed entries won't
be added back to the JBM_POSTOFFICE table.

So if you want the node re-join the cluster without restart, you need to set KeepOldFailoverModel to false and 
set NodeStateRefreshInterval to a reasonable value on both nodes. When you simulate the network issue, make sure
the time between 'ifdown' and 'ifup' is within the safe range (about 2*NodeStateRefreshInterval) so that the rejoining
can be done successfully.

In any case, the entries for local clustered queues will always in the table. I haven't seen them all gone in the tests.


                
> Clustered queues get deleted from the jbm_postoffice table when it should not
> -----------------------------------------------------------------------------
>
>                 Key: JBMESSAGING-1945
>                 URL: https://issues.jboss.org/browse/JBMESSAGING-1945
>             Project: JBoss Messaging
>          Issue Type: Bug
>          Components: Messaging Core
>    Affects Versions: 1.4.8.SP9
>         Environment: SOA 5.2
>            Reporter: Tom Ross
>            Assignee: Yong Hao Gao
>
> After a JBoss Messaging cluster falls apart, the empty queues belonging to the node that has left the cluster are removed from the jbm_postoffice table. Under certain conditions even the empty clustered queues belonging to the remaining node are removed as well

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira