[
https://issues.apache.org/jira/browse/HBASE-3891?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13054543#comment-13054543
]
Ted Yu commented on HBASE-3891:
-------------------------------
HRegion.compact() keeps a reference to the proxy returned by
TaskMonitor.get().createStatus()
If MonitoredTaskImpl@51bfa303 corresponds to this proxy, I don't know why
weakProxy.get() returned null.
> TaskMonitor is used wrong in some places
> ----------------------------------------
>
> Key: HBASE-3891
> URL: https://issues.apache.org/jira/browse/HBASE-3891
> Project: HBase
> Issue Type: Bug
> Components: regionserver
> Affects Versions: 0.92.0
> Reporter: Lars George
> Fix For: 0.92.0
>
>
> I have a long running log replay in progress but none of the updates show.
> This is caused by reusing the MonitorTask references wrong, and manifests
> itself like this in the logs:
> {noformat}
> 2011-05-16 15:22:18,127 WARN org.apache.hadoop.hbase.monitoring.TaskMonitor:
> Status org.apache.hadoop.hbase.monitoring.MonitoredTaskImpl@51bfa303 appears
> to have been leaked
> 2011-05-16 15:22:18,128 DEBUG
> org.apache.hadoop.hbase.monitoring.MonitoredTask: cleanup.
> {noformat}
> The cleanup sets the completion timestamp and causes the task to be purged
> from the list. After that the UI for example does not show any further
> running tasks, although from the logs I can see (with my log additions):
> {noformat}
> 2011-05-16 15:29:52,296 DEBUG
> org.apache.hadoop.hbase.monitoring.MonitoredTask: setStatus: Compaction
> complete: 103.1m in 18542ms
> 2011-05-16 15:29:52,296 DEBUG
> org.apache.hadoop.hbase.monitoring.MonitoredTask: setStatus: Running
> coprocessor post-compact hooks
> 2011-05-16 15:29:52,296 DEBUG
> org.apache.hadoop.hbase.monitoring.MonitoredTask: setStatus: Compaction
> complete
> 2011-05-16 15:29:52,297 DEBUG
> org.apache.hadoop.hbase.monitoring.MonitoredTask: markComplete: Compaction
> complete
> {noformat}
> They are silently ignored as the TaskMonitor has dropped their reference. We
> need to figure out why a supposedly completed task monitor was reused.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
