[
https://issues.apache.org/jira/browse/YARN-1029?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13850087#comment-13850087
]
Bikas Saha commented on YARN-1029:
----------------------------------
I am not sure what additional work is needed in YARN-1177 given that the RM
already implements the HAServiceProtocol. There shouldnt be any since the API
needed for ZKFC to manage the RM are already there. Perhaps we only need some
logic in the RM for some additional checks during automatic failover. So
functionally ZKFC supports should already be there.
This jira was meant to add built-in failover in the RM given that RM already
uses ZK store and to simplify deployments. I think our initial guess was the
ZKFC would be simpler to embed but Karthiks observation after writing some code
is that the elector library is simpler to embed. Lets review the patches to see
the merits in the code.
I agree that we dont need to merge the sessions for leader election and storage
right now. They are notionally separate and lets observe a case where one
session gets lost but not the other before trying to merge them together.
> Allow embedding leader election into the RM
> -------------------------------------------
>
> Key: YARN-1029
> URL: https://issues.apache.org/jira/browse/YARN-1029
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Bikas Saha
> Assignee: Karthik Kambatla
> Attachments: embedded-zkfc-approach.patch, yarn-1029-0.patch,
> yarn-1029-0.patch, yarn-1029-approach.patch
>
>
> It should be possible to embed common ActiveStandyElector into the RM such
> that ZooKeeper based leader election and notification is in-built. In
> conjunction with a ZK state store, this configuration will be a simple
> deployment option.
--
This message was sent by Atlassian JIRA
(v6.1.4#6159)
