[
https://issues.apache.org/jira/browse/HADOOP-1771?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
lohit vijayarenu updated HADOOP-1771:
-------------------------------------
Attachment: h-1771-3.patch
Thanks to Koji, while debugging a similar problem, we saw similar case where
MROutputThread was done, but the map task was still hung
Upon killing the streaming map executable, we saw there was a problem with the
thread so it had terminated.
Looks like the streaming map executable was still trying to write and was hung.
The problem could be that clientIn_ and clientErr_ are not closed when their
thread is done, which causes this problem.
Attached is an updated patch, which makes sure it will close the streams when
their threads exit.
Please could anyone review/comment.
Thanks.
> streaming hang when IOException in MROutputThread. (NPE)
> --------------------------------------------------------
>
> Key: HADOOP-1771
> URL: https://issues.apache.org/jira/browse/HADOOP-1771
> Project: Hadoop
> Issue Type: Bug
> Components: contrib/streaming
> Affects Versions: 0.13.1
> Reporter: Koji Noguchi
> Assignee: Koji Noguchi
> Priority: Minor
> Fix For: 0.15.0
>
> Attachments: h-1771-2.patch, h-1771-3.patch, h-1771.patch
>
>
> One streaming task hang and had stderr userlog as follows.
> {code}
> Exception in thread "Thread-5" java.lang.NullPointerException
> at java.lang.Throwable.printStackTrace(Throwable.java:460)
> at
> org.apache.hadoop.streaming.PipeMapRed$MROutputThread.run(PipeMapRed.java:352)
> {code}
> In PipeMapRed.java
> {code}
> 351 } catch (IOException io) {
> 352 io.printStackTrace(log_);
> 353 outerrThreadsThrowable = io;
> {code}
> I guess log_ is Null... Should call logStackTrace.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
