[ https://issues.apache.org/jira/browse/HBASE-14522?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14936796#comment-14936796 ]
Lars Francke commented on HBASE-14522: -------------------------------------- I tried it on my Mac and the Master process is somehow using lots of CPU when I start it with that flag. Haven't tried on Linux yet and will try to find out where this happens. > Document and/or disable hsperfdata GC pauses > -------------------------------------------- > > Key: HBASE-14522 > URL: https://issues.apache.org/jira/browse/HBASE-14522 > Project: HBase > Issue Type: Task > Components: documentation, Performance > Reporter: Nick Dimiduk > Assignee: Lars Francke > > {quote} > The JVM by default exports statistics by mmap-ing a file in /tmp > (hsperfdata). On Linux, modifying a memory mapped file can block until disk > I/O completes, which can be hundreds of milliseconds. Since the JVM modifies > these statistics during garbage collection and safepoints, this causes pauses > that are hundreds of milliseconds long. > {quote} > Via [JVM mmap pause|http://www.evanjones.ca/jvm-mmap-pause.html]. > We should add {{-XX:+PerfDisableSharedMem}} to our default options as was > apparently done in CASSANDRA-9242 and/or document the presence of this bug so > operators know to use tmpfs. > Hat-tip [~vrodionov] -- This message was sent by Atlassian JIRA (v6.3.4#6332)