hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jimmy Xiang (JIRA)" <j...@apache.org>
Subject [jira] [Created] (HBASE-11197) Region could remain unassigned if regionserver crashes
Date Sat, 17 May 2014 00:36:19 GMT
Jimmy Xiang created HBASE-11197:

             Summary: Region could remain unassigned if regionserver crashes
                 Key: HBASE-11197
                 URL: https://issues.apache.org/jira/browse/HBASE-11197
             Project: HBase
          Issue Type: Bug
          Components: Region Assignment
            Reporter: Jimmy Xiang
            Assignee: Jimmy Xiang

When looking into test failure: testVisibilityLabelsOnKillingOfRSContainingLabelsTable

and find this is what has happened:

1. try to assign a region a region server;
2. master creates a znode, and send an openRegion request to the rs;
3. rs gets the request and sends back a response, then crashed;
4. try to assign the region again with forceNewPlan = true;
5. since the region is in transition, master tries to close it and get region server stopped
6. master offlines the region and removes it from transition; but can't assign the region
since the dead server is not processed;
7. now SSH finally kicks in, tries to assign this region again;
8. SSH will fail to assign it since the znode is there already.

We should clean up the znode in force offline a region.

This message was sent by Atlassian JIRA

View raw message