You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues-all@impala.apache.org by "Tim Armstrong (JIRA)" <ji...@apache.org> on 2018/04/30 23:10:00 UTC
[jira] [Commented] (IMPALA-6857) Add JVM Pause Monitor to Impala
Processes
[ https://issues.apache.org/jira/browse/IMPALA-6857?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16459239#comment-16459239 ]
Tim Armstrong commented on IMPALA-6857:
---------------------------------------
Are there any JVM metrics about GC we could also expose? E.g. number of GCs, time of last GC, GC duration?
> Add JVM Pause Monitor to Impala Processes
> -----------------------------------------
>
> Key: IMPALA-6857
> URL: https://issues.apache.org/jira/browse/IMPALA-6857
> Project: IMPALA
> Issue Type: Improvement
> Components: Catalog, Frontend
> Reporter: Philip Zeyliger
> Priority: Major
> Labels: ramp-up, supportability
>
> In IMPALA-3114, we added a pause monitor for Impala. In addition to that, we should port/borrow Hadoop's JvmPauseMonitor [https://github.com/apache/hadoop/blob/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/util/JvmPauseMonitor.java.] I believe that when the JVM is aggressively GCing, the C++ threads will continue to get scheduled (and won't log), but the Java ones will log. (I've definitely seen JvmPauseMonitor be accurate many times.)
> [~bharathv], when you were testing this, were you able to reproduce it triggering when the JVM half was in "GC hell"?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-all-unsubscribe@impala.apache.org
For additional commands, e-mail: issues-all-help@impala.apache.org