[
https://issues.apache.org/jira/browse/YARN-1041?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13868169#comment-13868169
]
Vinod Kumar Vavilapalli commented on YARN-1041:
-----------------------------------------------
More comments on the patch:
- FifoScheduler depends on a concurrentSkipList. So you can't put the
initiation in the base class. You can leave the declaration though.
- There are some long lines in the patch.
- We don't need two APIs in RegResponse.newInstance(). They are not supposed
to be used by users.
- We don't need the explicit state-check in the method in the base scheduler
class.
> Protocol changes for RM to bind and notify a restarted AM of existing
> containers
> --------------------------------------------------------------------------------
>
> Key: YARN-1041
> URL: https://issues.apache.org/jira/browse/YARN-1041
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: resourcemanager
> Affects Versions: 3.0.0
> Reporter: Steve Loughran
> Assignee: Jian He
> Attachments: YARN-1041.1.patch, YARN-1041.2.patch
>
>
> For long lived containers we don't want the AM to be a SPOF.
> When the RM restarts a (failed) AM, it should be given the list of containers
> it had already been allocated. the AM should then be able to contact the NMs
> to get details on them. NMs would also need to do any binding of the
> containers needed to handle a moved/restarted AM.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
