[
https://issues.apache.org/jira/browse/YARN-6706?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16083261#comment-16083261
]
Arun Suresh commented on YARN-6706:
-----------------------------------
bq. You are referring to the else statement in
ContainerScheduler#scheduleContainer() right? The reason of doing
startPendingContainers() both before and after en-queuing is so that we always
respect the max queue limit for OPPORTUNISTIC containers.
Yup.. makes sense..
But was thinking about Karthik's
[comment|https://issues.apache.org/jira/browse/YARN-6706?focusedCommentId=16080813&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16080813]
and maybe we can get rid of *maxOppQueueLength*. I prefer just using the
*queuingLimit*.. It also reduces configuration hassle, since *queuingLimit* is
calculated centrally by the RM. The only reason we might need to have a locally
configured maxOppQueueLength is in the case of heterogeneous clusters where max
number of queued containers should differ from NM to NM - but to be honest, I
am not sure it would be useful.
> Refactor ContainerScheduler to make oversubscription change easier
> ------------------------------------------------------------------
>
> Key: YARN-6706
> URL: https://issues.apache.org/jira/browse/YARN-6706
> Project: Hadoop YARN
> Issue Type: Sub-task
> Affects Versions: 3.0.0-alpha3
> Reporter: Haibo Chen
> Assignee: Haibo Chen
> Attachments: YARN-6706.01.patch, YARN-6706-YARN-1011.00.patch,
> YARN-6706-YARN-1011.01.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
