hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "stack (JIRA)" <j...@apache.org>
Subject [jira] Updated: (HBASE-3331) Kill -STOP of RS hosting META does not recover
Date Fri, 10 Dec 2010 21:41:02 GMT

     [ https://issues.apache.org/jira/browse/HBASE-3331?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

stack updated HBASE-3331:

Bringing it into 0.90.0 so I can take a look at it. This case should work.  The server shutdown
knows which server had .META. and will not try scanning .META. to find down'd servers regions
if the server hosting .META. is the one we are processing.   So need to see whats up here.
 Jon has said in the past that we should use our in-memory state rather than go to .META.
but thats awkward to do IIRC given current structures.  Let me look at that again.

> Kill -STOP of RS hosting META does not recover
> ----------------------------------------------
>                 Key: HBASE-3331
>                 URL: https://issues.apache.org/jira/browse/HBASE-3331
>             Project: HBase
>          Issue Type: Bug
>    Affects Versions: 0.90.0
>            Reporter: Todd Lipcon
>            Priority: Critical
>         Attachments: timeouts.log.txt
> If you find the server hosting META and kill -STOP its region server, it will eventually
lose its ZK session and the master will split its logs and try to reassign. However, at some
point along here it tries to access the old META, and gets SocketTimeoutExceptions, which
cause it to keep retrying forever. Once I kill -9ed the stopped server, things came back to

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message