hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Joydeep Sen Sarma (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HDFS-1478) unsafe interrupted condition check in DFSClient lease renewal thread
Date Tue, 26 Oct 2010 21:33:23 GMT

    [ https://issues.apache.org/jira/browse/HDFS-1478?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12925148#action_12925148
] 

Joydeep Sen Sarma commented on HDFS-1478:
-----------------------------------------

see: https://issues.apache.org/bugzilla/show_bug.cgi?id=44157

with the patch for this - log4j is setting interrupted state for threads. i think this bug
and the comments suggest that there may be cases where InterruptedException is still being
propagated from log4j (which is actually the more likely culprit for MAPREDUCE-2157. so it
makes sense, as a precaution, to check for shutdown conditions inside interruptedexception
handlers.

> unsafe interrupted condition check in DFSClient lease renewal thread
> --------------------------------------------------------------------
>
>                 Key: HDFS-1478
>                 URL: https://issues.apache.org/jira/browse/HDFS-1478
>             Project: Hadoop HDFS
>          Issue Type: Bug
>            Reporter: Joydeep Sen Sarma
>            Assignee: dhruba borthakur
>
> the lease renewal thread bails out if thread interrupted condition is set:
>     while (clientRunning && !Thread.interrupted()) {
> this seems incorrect. we have seen cases (MAPREDUCE-2157) where (apparently) log4j may
be setting the interrupted condition. if that is the case - the lease renewal thread will
terminate - and i think we have heard of cases where lease renewal is not happening. so these
seem to fit.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message