[
https://issues.apache.org/jira/browse/HADOOP-2756?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12575136#action_12575136
]
Raghu Angadi commented on HADOOP-2756:
--------------------------------------
> The streamer thread might not exit (hence the streamer.join() might block
> indefinitely) if the reponseder thread is not closed (via response.close()).
I don't see dataStreamer waiting for repsonse to close(). If it is not closed,
it closes it.
I think attached patch is fine. I checked the peripheral code again and I don't
see it introducing any new problem and fixes this bug. Does anyone see why it
is wrong? If this fix is ok, I will submit a slightly reformatted patch (i.e.
bring all the references to 'response' inside closeInternal() to one place).
> NPE in DFSClient in hbase under load
> ------------------------------------
>
> Key: HADOOP-2756
> URL: https://issues.apache.org/jira/browse/HADOOP-2756
> Project: Hadoop Core
> Issue Type: Bug
> Components: dfs
> Affects Versions: 0.16.0
> Reporter: stack
> Assignee: Raghu Angadi
> Priority: Minor
> Fix For: 0.16.1
>
> Attachments: HADOOP-2756.patch
>
>
> Saw this in logs:
> {code}
> 2008-01-31 18:55:02,128 ERROR org.apache.hadoop.hbase.HRegionServer:
> Compaction failed for region TestTable,0009438931,1201805282651
> java.lang.NullPointerException
> at
> org.apache.hadoop.dfs.DFSClient$DFSOutputStream.close(DFSClient.java:2262)
> at
> org.apache.hadoop.fs.FSDataOutputStream$PositionCache.close(FSDataOutputStream.java:51)
> at
> org.apache.hadoop.fs.FSDataOutputStream.close(FSDataOutputStream.java:67)
> at org.apache.hadoop.hbase.HStoreFile.writeInfo(HStoreFile.java:365)
> at org.apache.hadoop.hbase.HStore.compact(HStore.java:1236)
> at org.apache.hadoop.hbase.HRegion.compactStores(HRegion.java:775)
> at org.apache.hadoop.hbase.HRegion.compactIfNeeded(HRegion.java:707)
> at
> org.apache.hadoop.hbase.HRegionServer$CompactSplitThread.run(HRegionServer.java:253)
> {code}
> Look to see if the response data method needs to be made volatile (There's a
> test for null just before we use it on line #2262).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
