[jira] [Commented] (ARTEMIS-2077) Split-Brain Resolution When Connection Is Regained

["Justin Bertram (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ARTEMIS-2077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16607166#comment-16607166 ] 

Justin Bertram commented on ARTEMIS-2077:
-----------------------------------------

Off the top of my head I think the simplest way to implement this would be to add a setting whereby {{org.apache.activemq.artemis.core.cluster.DiscoveryGroup#checkUniqueID}} would track the number of duplicate broadcasts and shutdown the broker if it exceeded a certain amount (dictated by the new setting).  The broadcast would need to add a new piece of data (e.g. server up-time) so an intelligent decision could be made about which broker should shutdown otherwise both brokers could end up shutting themselves down.

That said, it's really best not to get into this situation in the first place.  Have you employed any of the recommended means to prevent split brain in the first place (i.e. odd number of live/backup pairs in the cluster or network health checker)?

> Split-Brain Resolution When Connection Is Regained
> --------------------------------------------------
>
>                 Key: ARTEMIS-2077
>                 URL: https://issues.apache.org/jira/browse/ARTEMIS-2077
>             Project: ActiveMQ Artemis
>          Issue Type: Bug
>          Components: Broker
>            Reporter: Ilkka Virolainen
>            Priority: Major
>             Fix For: 2.6.2
>
>
> When the master node in a replicating master/slave pair is congested or isolated, a failover occurs resulting in split brain. As the cluster connection is regained, failback should occur to mitigate the situation.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)