[
https://issues.apache.org/jira/browse/FLINK-9960?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16556059#comment-16556059
]
Julio Biason commented on FLINK-9960:
-------------------------------------
Just to add: I tried to configure high-availability.storageDir to have the list
of servers in the format
"hdfs://namenode1:8020;namenode2:8020/flink/jobmanager", following the example
used for journal configuration in Hadoop itself and it didn't work.
> JobManager can't failover to hdfs with multiple namenodes
> ---------------------------------------------------------
>
> Key: FLINK-9960
> URL: https://issues.apache.org/jira/browse/FLINK-9960
> Project: Flink
> Issue Type: Bug
> Components: JobManager
> Affects Versions: 1.4.0
> Reporter: Julio Biason
> Priority: Major
>
> When Hadoop/HDFS is running in HA mode, there may be more than one namenode
> running: One active and another standby.
> If, for whatever reasons, the server configured in
> high-availability.storageDir is currently in standby mode (e.g., Hadoop
> failed over the other namenode), the JobManager will refuse to start because
> it can't contact hdfs – in this case, it should have a list of servers (up to
> 2) and, in case of a failure to connect, go to the next one.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
