[
https://issues.apache.org/jira/browse/YARN-4212?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15839007#comment-15839007
]
ASF GitHub Bot commented on YARN-4212:
--------------------------------------
Github user flyrain commented on a diff in the pull request:
https://github.com/apache/hadoop/pull/181#discussion_r97919802
--- Diff:
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/QueueManager.java
---
@@ -272,6 +272,16 @@ private FSQueue createNewQueues(FSQueueType queueType,
FSParentQueue newParent = null;
String queueName = i.next();
+ // Check if child policy is allowed
--- End diff --
Yes, my original thought is to do that in another JIRA. The depth and
parent-child policy are not the same. It mighty a good idea to combine them
since the logic of depth checking only prevent fifo policy to be non-leaf
queue. The current implementation seems a bit heavy. I can do it in this JIRA.
> FairScheduler: Parent queues is not allowed to be 'Fair' policy if its
> children have the "drf" policy
> -----------------------------------------------------------------------------------------------------
>
> Key: YARN-4212
> URL: https://issues.apache.org/jira/browse/YARN-4212
> Project: Hadoop YARN
> Issue Type: Bug
> Reporter: Arun Suresh
> Assignee: Yufei Gu
> Labels: fairscheduler
> Attachments: YARN-4212.002.patch, YARN-4212.003.patch,
> YARN-4212.004.patch, YARN-4212.005.patch, YARN-4212.006.patch,
> YARN-4212.007.patch, YARN-4212.1.patch
>
>
> The Fair Scheduler, while performing a {{recomputeShares()}} during an
> {{update()}} call, uses the parent queues policy to distribute shares to its
> children.
> If the parent queues policy is 'fair', it only computes weight for memory and
> sets the vcores fair share of its children to 0.
> Assuming a situation where we have 1 parent queue with policy 'fair' and
> multiple leaf queues with policy 'drf', Any app submitted to the child queues
> with vcore requirement > 1 will always be above fairshare, since during the
> recomputeShare process, the child queues were all assigned 0 for fairshare
> vcores.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
