[
https://issues.apache.org/jira/browse/HADOOP-13796?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15644375#comment-15644375
]
Kihwal Lee commented on HADOOP-13796:
-------------------------------------
The failure detection works for the scenario you described. However, failover
may not work unless the fencing is configured properly.
Since the graceful failover won't work, the fencing will kick in to make sure
the failed node won't come back and act as an active. If you haven't configured
a proper fencing method, fail-over won't happen. E.g. if the only configured
fencing method is ssh-based, it won't work for this case. To fence off a
completely unresponsive node, the fencing script should be capable of cutting
the power.
> Standby namenode didnot change its role
> ---------------------------------------
>
> Key: HADOOP-13796
> URL: https://issues.apache.org/jira/browse/HADOOP-13796
> Project: Hadoop Common
> Issue Type: Bug
> Components: common
> Affects Versions: 2.6.4
> Environment: testing
> Reporter: Bhavan Jindal
>
> I have a high available apache Hadoop cluster using zookeeper quorum method.
> where zookeeper is monitoring the active namenode.
> When the daemon goes down, it automatically flips the role fine.
> However when the machine dies (complete halt), it didnot flip the roles.
> rather waited for the machine to comeup and then flip the roles.
> Is it an expected behavior OR design fault. i am using hadoop 2.6.4
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
