[
https://issues.apache.org/jira/browse/HDFS-1696?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13001902#comment-13001902
]
Suresh Srinivas commented on HDFS-1696:
---------------------------------------
More description of the change:
In federation block pools are upgraded one at a time. Consider a case where a
datanode is started with a later release and the namenode is not yet upgraded.
In that case, instead of shutting down or stop talking to the namenode, the
datanode continues to periodically perform handshake with the namenode. This
ensures that if one or more namenodes are not upgraded yet, the datanode
continues to perform handshakes and finally upgrades the block pool
corresponding to the namenode, when the namenode is upgraded.
This helps need for restarting the datanode, if a namenode had not been
upgraded when the cluster started up.
> HDFS Federation: when build version doesn't match - datanode should wait
> (keep connecting) untill NN comes up with the right version
> -------------------------------------------------------------------------------------------------------------------------------------
>
> Key: HDFS-1696
> URL: https://issues.apache.org/jira/browse/HDFS-1696
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: name-node
> Reporter: Boris Shkolnik
> Assignee: Boris Shkolnik
> Attachments: HDFS-1696-1.patch, HDFS-1696.patch
>
>
--
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
