You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Todd Lipcon (JIRA)" <ji...@apache.org> on 2010/04/14 22:54:48 UTC
[jira] Updated: (HBASE-2448) Scanner threads are interrupted
without acquiring lock properly
[ https://issues.apache.org/jira/browse/HBASE-2448?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Todd Lipcon updated HBASE-2448:
-------------------------------
Attachment: hbase-2448.txt
Here's a patch which changes around the interruptions to only trigger the sleeper to end its sleep early (or skip its next sleep cycle)
To test, I set my meta scan frequency to 1ms so that the HMaster is constantly scanning. Without this patch it ends up crashing the master pretty quickly when regions are reassigned, etc. With the patch I managed to run load test for some time without seeing it reoccur.
> Scanner threads are interrupted without acquiring lock properly
> ---------------------------------------------------------------
>
> Key: HBASE-2448
> URL: https://issues.apache.org/jira/browse/HBASE-2448
> Project: Hadoop HBase
> Issue Type: Bug
> Components: master
> Reporter: Todd Lipcon
> Assignee: Todd Lipcon
> Priority: Critical
> Attachments: hbase-2448.txt
>
>
> There are a few places where scanner threads are interrupted with .interrupt() instead of .interruptIfAlive(). This means that if they're in the midst of the checkFileSystem operation, it'll end up catching the interruption there, determine that the filesystem is down, and shut down the whole server. Other nasties can also result.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira