[
https://issues.apache.org/jira/browse/HBASE-2758?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12881528#action_12881528
]
Jonathan Gray commented on HBASE-2758:
--------------------------------------
Patch looks good. This also fixes an existing race condition where the master
node in ZK was put up before the master got the listing of regionservers.
Nothing would be blocking the RS from putting up their ephemeral nodes so it
was possible that an HMaster thought it was a failover but it was a clean
startup. Test added in patch verifies that a cluster will startup even if
there are unassigned znodes in zookeeper.
Running full test suite and then will commit.
> META region stuck in RS2ZK_REGION_OPENED state
> ----------------------------------------------
>
> Key: HBASE-2758
> URL: https://issues.apache.org/jira/browse/HBASE-2758
> Project: HBase
> Issue Type: Bug
> Components: master, regionserver
> Affects Versions: 0.21.0
> Reporter: Todd Lipcon
> Assignee: Karthik Ranganathan
> Priority: Blocker
> Attachments: HBASE-2758-0.21.patch
>
>
> In cluster testing trunk, I ended up with a situation where META was
> unassigned and no amount of restarting various pieces would fix it. On master
> startup, I see:
> 2010-06-20 21:08:05,431 DEBUG org.apache.hadoop.hbase.master.BaseScanner:
> Current assignment of .META.,,1.1028785192 is not valid; serverAddress=,
> startCode=0 unknown.
> 2010-06-20 21:08:05,436 DEBUG
> org.apache.hadoop.hbase.zookeeper.ZooKeeperWrapper: While creating UNASSIGNED
> region 1028785192 exists, state = RS2ZK_REGION_OPENED
> 2010-06-20 21:08:05,438 WARN
> org.apache.hadoop.hbase.zookeeper.ZooKeeperWrapper:
> <monster01.sf.cloudera.com:/hbase,org.apache.hadoop.hbase.master.HMaster>Failed
> to create ZNode /hbase/UNASSIGNED/1028785192 in ZooKeeper
> org.apache.zookeeper.KeeperException$NodeExistsException: KeeperErrorCode =
> NodeExists for /hbase/UNASSIGNED/1028785192
> 2010-06-20 21:08:05,438 DEBUG org.apache.hadoop.hbase.master.RegionManager:
> Created UNASSIGNED zNode .META.,,1.1028785192 in state M2ZK_REGION_OFFLINE
> then on the RS:
> 2010-06-20 21:08:05,899 ERROR
> org.apache.hadoop.hbase.regionserver.RSZookeeperUpdater: ZNode
> /hbase/UNASSIGNED/1028785192 is not in CLOSED/OFFLINE state (state =
> RS2ZK_REGION_OPENED), will NOT open region.
> 2010-06-20 21:08:05,899 ERROR
> org.apache.hadoop.hbase.regionserver.HRegionServer: Error opening
> .META.,,1.1028785192
> java.io.IOException: ZNode /hbase/UNASSIGNED/1028785192 is not in
> CLOSED/OFFLINE state (state = RS2ZK_REGION_OPENED), will NOT open region.
> and the region never opens
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
