[
https://issues.apache.org/jira/browse/HADOOP-16248?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexis Daboville updated HADOOP-16248:
--------------------------------------
Description:
In some circumstances (high GC, high CPU usage, creating lots of
S3AFileSystem) it is possible for MutableQuantiles::scheduler [1] to fall
behind processing tasks that are submitted to it; because tasks are
submitted on a regular schedule, the unbounded queue backing the
{{ExecutorService}} might grow to several gigs [2]. By using
{{scheduleWithFixedDelay}} instead, we ensure that under pressure this leak
won't
happen. In order to mitigate the growth, a simple fix [3] is proposed, simply
replacing {{scheduler.scheduleAtFixedRate}} by
{{scheduler.scheduleWithFixedDelay}}.
[1] it is single threaded and shared across all instances of
{{MutableQuantiles}}:
[https://github.com/apache/hadoop/blob/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics2/lib/MutableQuantiles.java#L66-L68]
[2] see attached mutable-quantiles-leak.png.
[3] mutable-quantiles.patch
was:
In some circumstances (high GC, high CPU usage, creating lots of
S3AFileSystem) it is possible for MutableQuantiles::scheduler [1] to fall
behind processing tasks that are submitted to it; because tasks are
submitted on a regular schedule, the unbounded queue backing the
{{ExecutorService}} might grow to several gigs [2]. By using
{{scheduleWithFixedDelay}} instead, we ensure that under pressure this leak
won't
happen. In order to mitigate the growth, a simple fix is proposed, simply
replacing {{scheduler.scheduleAtFixedRate}} by
{{scheduler.scheduleWithFixedDelay}}.
[1] it is single threaded and shared across all instances of
{{MutableQuantiles}}:
[https://github.com/apache/hadoop/blob/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics2/lib/MutableQuantiles.java#L66-L68]
[2] see attached mutable-quantiles-leak.png.
> Fix MutableQuantiles memory leak
> --------------------------------
>
> Key: HADOOP-16248
> URL: https://issues.apache.org/jira/browse/HADOOP-16248
> Project: Hadoop Common
> Issue Type: Bug
> Affects Versions: 2.9.2
> Reporter: Alexis Daboville
> Priority: Major
> Attachments: mutable-quantiles-leak.png, mutable-quantiles.patch
>
>
> In some circumstances (high GC, high CPU usage, creating lots of
> S3AFileSystem) it is possible for MutableQuantiles::scheduler [1] to fall
> behind processing tasks that are submitted to it; because tasks are
> submitted on a regular schedule, the unbounded queue backing the
> {{ExecutorService}} might grow to several gigs [2]. By using
> {{scheduleWithFixedDelay}} instead, we ensure that under pressure this leak
> won't
> happen. In order to mitigate the growth, a simple fix [3] is proposed,
> simply replacing {{scheduler.scheduleAtFixedRate}} by
> {{scheduler.scheduleWithFixedDelay}}.
> [1] it is single threaded and shared across all instances of
> {{MutableQuantiles}}:
> [https://github.com/apache/hadoop/blob/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics2/lib/MutableQuantiles.java#L66-L68]
> [2] see attached mutable-quantiles-leak.png.
> [3] mutable-quantiles.patch
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
