[ https://issues.apache.org/jira/browse/HADOOP-7256?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Harsh J updated HADOOP-7256:
----------------------------
Affects Version/s: (was: 0.21.0)
> Resource leak during failure scenario of closing of resources.
> ---------------------------------------------------------------
>
> Key: HADOOP-7256
> URL: https://issues.apache.org/jira/browse/HADOOP-7256
> Project: Hadoop Common
> Issue Type: Bug
> Affects Versions: 0.20.2
> Reporter: ramkrishna.s.vasudevan
> Priority: Minor
> Attachments: HADOOP-7256.patch, HADOOP-7256-patch-1.patch, HADOOP-7256-patch-2.patch
>
> Original Estimate: 8h
> Remaining Estimate: 8h
>
> Problem Statement:
> ===============
> There are chances of resource leak and stream not getting closed
> Take the case when after copying data we try to close the Input and output stream followed
by closing of the socket.
> Suppose an exception occurs while closing the input stream(due to runtime exception)
then the subsequent operations of closing the output stream and socket may not happen and
there is a chance of resource leak.
> Scenario
> =======
> During long run of map reduce jobs, the copyFromLocalFile() api is getting called.
> Here we found some exceptions happening. As a result of this we found the lsof value
raising leading to resource leak.
> Solution:
> =======
> While doing a close operation of any resource catch the RuntimeException also rather
than catching the IOException alone.
> Additionally there are places where we try to close a resource in the catch block.
> If this close fails, we just throw and come out of the current flow.
> In order to avoid this, we can carry out the close operation in the finally block.
> Probable reasons for getting RunTimeExceptions:
> =====================================
> We may get runtime exception from customised hadoop streams like FSDataOutputStream.close()
. So better to handle RunTimeExceptions also.
>
--
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
|