[
https://issues.apache.org/jira/browse/HBASE-7981?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13595371#comment-13595371
]
Himanshu Vashishtha commented on HBASE-7981:
--------------------------------------------
I think it is the way it looks for the daughter regions in the meta table is
buggy: the regionname() it uses to navigate the scan in the meta scanner makes
it work like the region doesn't exist and then it creates a region for that.
For example, in the above example:
{code}
************* 345c528e4c394953bca5695d4774daef *******
First split:
******** e74872f8b6cd6edc95f8c2b2d7924fa5 **********
38ba9b561bbf6db5e172ee2118fc2cff *******
Second split:
***e74872f8b6cd6edc95f8c2b2d7924fa5 *** 914d5913260dc2e1698a523001905127 ***
bdb74bd7defe897e6277e476b47e9b56 ***
{code}
The servershutdownhandler starts to look for the daughter regions and so on. It
is able to find the very first region in the meta:
345c528e4c394953bca5695d4774daef. Then, it looks for its daughters. It found
the first one.
{code}
2013-03-05 14:24:47,262 DEBUG
[MASTER_SERVER_OPERATIONS-janus.apache.org,41462,1362493469668-2]
handler.ServerShutdownHandler(375): Daughter testShutdownFixupWh
enDaughterHasSplit,,1362493482343.e74872f8b6cd6edc95f8c2b2d7924fa5. present
{code}
Then it looks for the second one:
{code}
2013-03-05 14:24:47,262 DEBUG
[MASTER_SERVER_OPERATIONS-janus.apache.org,41462,1362493469668-2]
client.ClientScanner(96): Creating scanner over .META. starting at key
'testShutdownFixupWhenDaughterHasSplit,mnk,1362493482343.38ba9b561bbf6db5e172ee2118fc2cff.,,00000000000000'
{code}
It use the start key as
'testShutdownFixupWhenDaughterHasSplit,mnk,1362493482343.38ba9b561bbf6db5e172ee2118fc2cff.,,00000000000000'.
I don't know why it append that 00000000000 after wards, but it is not able to
find this daughter, and then adds it to meta.
This causes that we have one extra region in the meta in OPEN state. It doesn't
always append that 00000000 suffix.
{code}
2013-03-05 14:24:47,262 DEBUG
[MASTER_SERVER_OPERATIONS-janus.apache.org,41462,1362493469668-2]
client.ClientScanner(209): Advancing internal scanner to startKey at
'testShutdownFixupWhenDaughterHasSplit,mnk,1362493482343.38ba9b561bbf6db5e172ee2118fc2cff.,,00000000000000'
2013-03-05 14:24:47,265 DEBUG
[MASTER_SERVER_OPERATIONS-janus.apache.org,41462,1362493469668-2]
client.ClientScanner(196): Finished with scanning at {NAME => '.META.,,1',
STARTKEY => '', ENDKEY => '', ENCODED => 1028785192,}
2013-03-05 14:24:47,265 DEBUG
[MASTER_SERVER_OPERATIONS-janus.apache.org,41462,1362493469668-2]
hbase.HRegionInfo(964): Current INFO from scan results = {NAME =>
'testShutdownFixupWhenDaughterHasSplit,suc,1362493483680.914d5913260dc2e1698a523001905127.',
STARTKEY => 'suc', ENDKEY => '', ENCODED => 914d5913260dc2e1698a523001905127,}
2013-03-05 14:24:47,265 INFO
[MASTER_SERVER_OPERATIONS-janus.apache.org,41462,1362493469668-2]
handler.ServerShutdownHandler(364): Fixup; missing daughter test
ShutdownFixupWhenDaughterHasSplit,mnk,1362493482343.38ba9b561bbf6db5e172ee2118fc2cff.
{code}
> TestSplitTransactionOnCluster.testShutdownFixupWhenDaughterHasSplit failed in
> 0.95 build #11
> --------------------------------------------------------------------------------------------
>
> Key: HBASE-7981
> URL: https://issues.apache.org/jira/browse/HBASE-7981
> Project: HBase
> Issue Type: Bug
> Reporter: stack
> Assignee: stack
> Priority: Blocker
> Fix For: 0.95.0
>
> Attachments: 7981.extra.logging.txt, 7981.extra.logging.txt,
> 7981v2.extra.logging.txt, HBASE-7981.patch
>
>
> https://builds.apache.org/job/hbase-0.95/11/testReport/junit/org.apache.hadoop.hbase.regionserver/TestSplitTransactionOnCluster/testShutdownFixupWhenDaughterHasSplit/
> Hard to tell which region is missing post crash. Not logged.
--
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