[
https://issues.apache.org/jira/browse/YARN-5825?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15654156#comment-15654156
]
Sunil G edited comment on YARN-5825 at 11/10/16 2:19 PM:
---------------------------------------------------------
Attaching an initial version of patch.
In PCPP#cloneQueues, we use CSQueue abstract object. So I had to add
{{getReadLock}} as an api in CSQueue interface to do lock. I agree this is not
so clean way, however the alternative way is also not clean.
We might need to add below code in PCPP if we want to remove {{getReadQueue}}
from CSQueue interface
{code}
private ReentrantReadWriteLock.ReadLock getQueueReadLock(CSQueue curQueue) {
if (curQueue instanceof ParentQueue) {
return ((ParentQueue) curQueue).getReadLock();
} else if (curQueue instanceof LeafQueue) {
return ((LeafQueue) curQueue).getReadLock();
}
return null;
}
{code}
[~leftnoteasy] [~jianhe] thoughts?
was (Author: sunilg):
Attaching an initial version of patch.
In PCPP#cloneQueues, we use CSQueue abstract object. So I had to add
{{getReadLock}} as an api in CSQueue interface to do lock. I agree this is not
so clean way, however the alternative way is also not clean.
We might need to add below code in PCPP if we want to remove {{getReadQueue}}
from CSQueue interface
{code}
private ReentrantReadWriteLock.ReadLock getQueueReadLock(CSQueue curQueue) {
if (curQueue instanceof ParentQueue) {
return ((ParentQueue) curQueue).getReadLock();
} else if (curQueue instanceof LeafQueue) {
return ((LeafQueue) curQueue).getReadLock();
}
return null;
}
{code}
[~leftnoteasy] thoughts?
> ProportionalPreemptionalPolicy could use readLock over LeafQueue instead of
> synchronized block
> ----------------------------------------------------------------------------------------------
>
> Key: YARN-5825
> URL: https://issues.apache.org/jira/browse/YARN-5825
> Project: Hadoop YARN
> Issue Type: Bug
> Components: capacity scheduler
> Reporter: Sunil G
> Assignee: Sunil G
> Attachments: YARN-5825.0001.patch
>
>
> Currently in PCPP, {{synchronized (curQueue)}} is used in various places.
> Such instances could be replaced with a read lock. Thank you [~jianhe] for
> pointing out the same as comment
> [here|https://issues.apache.org/jira/browse/YARN-2009?focusedCommentId=15626578&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15626578]
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
