[
https://issues.apache.org/jira/browse/YARN-3091?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16244511#comment-16244511
]
Wangda Tan commented on YARN-3091:
----------------------------------
[~templedf]/[~jlowe],
The RW locks are introduced to solve multiple threads to look at container
allocation problem. From my tests report:
https://issues.apache.org/jira/secure/attachment/12831662/YARN-5139-Concurrent-scheduling-performance-report.pdf,
it can get about 2.5X throughput improvement when we have 3 threads looking at
scheduler at the same time comparing to single thread.
I agree that some previous locking changes (such as
YARN-3139/YARN-3140/YARN-3141) can definitely be improved. But I think if we
change everything to simple reentrant lock may affect throughput when we have
multiple threads to do allocation.
> [Umbrella] Improve and fix locks of RM scheduler
> ------------------------------------------------
>
> Key: YARN-3091
> URL: https://issues.apache.org/jira/browse/YARN-3091
> Project: Hadoop YARN
> Issue Type: Task
> Components: capacityscheduler, fairscheduler, resourcemanager,
> scheduler
> Reporter: Wangda Tan
>
> In existing YARN RM scheduler, there're some issues of using locks. For
> example:
> - Many unnecessary synchronized locks, we have seen several cases recently
> that too frequent access of scheduler makes scheduler hang. Which could be
> addressed by using read/write lock. Components include scheduler, CS queues,
> apps
> - Some fields not properly locked (Like clusterResource)
> We can address them together in this ticket.
> (More details see comments below)
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
