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-8957) TestDrainingServer. testDrainingServerWithAbort fails
Date Tue, 16 Jul 2013 21:36:50 GMT

     [ https://issues.apache.org/jira/browse/HBASE-8957?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

stack updated HBASE-8957:
-------------------------

    Attachment: 8957.txt

Issue is permissions when running with short-circuit reads.  Recovering of the .META. log
edits fails and so .META. is stuck in RIT till the test times out.

In the logs you see:
{code}
org.apache.hadoop.security.AccessControlException: Can't continue with getBlockLocalPathInfo()
authorization. The user jenkins.hfs.4 is not allowed to call getBlockLocalPathInfo
	at org.apache.hadoop.hdfs.server.datanode.DataNode.checkBlockLocalPathAccess(DataNode.java:1013)

{code}

The 'jenkins.hfs.4' is the RS trying to replay edits.  It is failing because only the user
'jenkins' has perms to do short circuit reading.

The '.hfs.4' suffix is added in HBaseTestingUtility when we start up a little hbase cluster...
it is added per regionserver.

So, we could reach back into the little mnincluster and IFF shortcircuit reads are enabled,
then we have all RS use same user accessing hdfs.

But I think this will break tests.  So I went the other route and disabled shortcircuit reading
as the default for mini clusters.  Quiet a few tests disable it already.

We probably need to add an explicit shortcircuit tests.

{code}
 30 Patch that disables shortcircuit reading as default when running unit tests
 29 because when a minicluster, not all regionservers have permissions to do
 28 short circuit reading.
 27
 26 May slow tests some.
 25
 24
 23 M hbase-client/src/main/java/org/apache/hadoop/hbase/zookeeper/ZKAssign.java
 22   Unrelated: Shorten log message and add some context to explain a mysterious log.
 21
 20 M hbase-server/src/main/java/org/apache/hadoop/hbase/master/AssignmentManager.java
 19   Unrelated: Shorten a log message.
 18
 17
 16 M hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java
 15   Unrelated refactoring of a getMetaWAL method.  This getMetaWAL is actually kinda
 14   wrong in that it will do the creation if does not already exist which makes for
 13   interesting parents:
 12
 11     2013-07-15 12:04:04,245 INFO  [RS_OPEN_META-juno:36201-0MetaLogRoller] regionserver.LogRoller(119):
LogRoller exiting.
 10
  9   An OPEN executor has the meta as its parent it seems which probably means
  8   the executor has an occupied thread, a thread that won't go down.  TODO.
  7
  6   Also fix name of metalogroller thread so it is like others.
  5
  4 M hbase-server/src/test/java/org/apache/hadoop/hbase/HBaseTestingUtility.java
  3   Main fix changing short circuit read to default off.
  2
  1 M hbase-server/src/test/java/org/apache/hadoop/hbase/TestDrainingServer.java
  0   Unrelated... add some sleep so we don't spew logs.
{code}
                
> TestDrainingServer. testDrainingServerWithAbort fails
> -----------------------------------------------------
>
>                 Key: HBASE-8957
>                 URL: https://issues.apache.org/jira/browse/HBASE-8957
>             Project: HBase
>          Issue Type: Bug
>            Reporter: stack
>            Assignee: stack
>         Attachments: 8957.txt
>
>
> See here https://builds.apache.org/job/hbase-0.95-on-hadoop2/179/testReport/org.apache.hadoop.hbase/TestDrainingServer/testDrainingServerWithAbort/
> and here
> http://54.241.6.143/job/HBase-0.95-Hadoop-2/org.apache.hbase$hbase-server/600/testReport/org.apache.hadoop.hbase/TestDrainingServer/testDrainingServerWithAbort/
> It looks like .META. is not being assigned.  Am digging in  here now.

--
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

Mime
View raw message