[
https://issues.apache.org/jira/browse/MAPREDUCE-5870?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14745511#comment-14745511
]
Jason Lowe commented on MAPREDUCE-5870:
---------------------------------------
I too struggle with whether it's worth going through all this trouble. I
honestly don't know how much the MRv1 priority API was used and therefore the
impact of completely abandoning it.
If we are going to keep supporting the MRv1 priority API, what are your
thoughts about handling the reporting of app priority via JobStatus from the
AM? I really don't want YARNRunner hitting the RM for every status update, as
that will put a lot more RPC load on the RM than we have today. But the
alternative means we either need to have the AM report a delayed view of the
priority (e.g.: send priority in heartbeat response, AM may be out-of-sync by
up to one heartbeat interval), or we have to deprecate that part of the MRv1
API.
I think I'm OK with having the priority reported in the AM heartbeat. It will
be somewhat delayed with reality when it changes, but it should be relatively
rare for it to be out of sync more than a few seconds. The priority would have
to have changed while the AM is having difficulty contacting the RM. If that
persists too long either the RM will expire the AM and kill it or the AM will
give up on the RM and exit. Having the AM know the priority could also help
with app frameworks that want to report this in their AM-provided UI.
> Support for passing Job priority through Application Submission Context in
> Mapreduce Side
> -----------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-5870
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5870
> Project: Hadoop Map/Reduce
> Issue Type: Improvement
> Components: client
> Reporter: Sunil G
> Assignee: Sunil G
> Attachments: 0001-MAPREDUCE-5870.patch, 0002-MAPREDUCE-5870.patch,
> 0003-MAPREDUCE-5870.patch, 0004-MAPREDUCE-5870.patch,
> 0005-MAPREDUCE-5870.patch, 0006-MAPREDUCE-5870.patch, Yarn-2002.1.patch
>
>
> Job Prioirty can be set from client side as below [Configuration and api].
> a. JobConf.getJobPriority() and
> Job.setPriority(JobPriority priority)
> b. We can also use configuration
> "mapreduce.job.priority".
> Now this Job priority can be passed in Application Submission
> context from Client side.
> Here we can reuse the MRJobConfig.PRIORITY configuration.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
