hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "rajeshbabu (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-8133) avoid assign for disabling table regions(OPENING/PENDING_OPEN) in SSH
Date Mon, 18 Mar 2013 04:46:16 GMT

    [ https://issues.apache.org/jira/browse/HBASE-8133?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13604868#comment-13604868

rajeshbabu commented on HBASE-8133:

bq. This change is needed but is this scenario possible?
This case usually happen with balance and disableTable race. Lets suppose a region is moving
from RS1 to RS2 and it is in OPENING/PENDING_OPEN then master recieved disable table call.
If RS2 went down after setting DISABLING state then we are unnessarily assigning and then
unassigning the regions in SSH. This can be easily reproducable. 

bq. Ideally i prefer we work on the test case to make it more accurate as per the working.
Yes,its needed. In TestMasterFailover we are setting table state to DISABLED and then we are
creating transitions. Ideally it should not happen because if table is disabled then the table
regions should never be transition or online. We need to find such cases if possible and fix
> avoid assign for disabling table regions(OPENING/PENDING_OPEN) in SSH
> ---------------------------------------------------------------------
>                 Key: HBASE-8133
>                 URL: https://issues.apache.org/jira/browse/HBASE-8133
>             Project: HBase
>          Issue Type: Bug
>          Components: Region Assignment
>    Affects Versions: 0.95.0
>            Reporter: rajeshbabu
>            Assignee: rajeshbabu
>             Fix For: 0.98.0
> Disabling table regions which are in PENDING_OPEN or OPENING on dead server are getting
assigned then unassiging. 
> We can avoid this by just mark OFFLINE for the regions,any way znodes of the transitions
got deleted as part of am.processServerShutdown(serverName). 

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

View raw message