[jira] Commented: (HADOOP-5009) DataNode#shutdown sometimes leaves data block scanner verification log unclosed

["Tsz Wo (Nicholas), SZE (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HADOOP-5009?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12663129#action_12663129 ] 

Tsz Wo (Nicholas), SZE commented on HADOOP-5009:
------------------------------------------------

+1 Patch looks good.  Two minor suggestions:
- using blockScannerThread.join() may be dangerous if blockScannerThread does not terminate.  How about replace join() with blockScannerThread.join(millis), so that it will time out?
- I think it is better to use try-finally in closing the readers.  See [a comment|https://issues.apache.org/jira/browse/HADOOP-2926?focusedCommentId=12574664#action_12574664] in HADOOP-2926.

> DataNode#shutdown sometimes leaves data block scanner verification log unclosed
> -------------------------------------------------------------------------------
>
>                 Key: HADOOP-5009
>                 URL: https://issues.apache.org/jira/browse/HADOOP-5009
>             Project: Hadoop Core
>          Issue Type: Bug
>          Components: dfs
>    Affects Versions: 0.17.0
>            Reporter: Hairong Kuang
>            Assignee: Hairong Kuang
>             Fix For: 0.19.1
>
>         Attachments: blockScannerFd.patch
>
>
> When datanode gets shutdown by calling DataNode#shutdown, it occasionally leaves the data block scanner verification log unclosed. There are two possible causes:
> 1. DataNode does not wait until block scanner thread to exit.
> 2. DataBlockScanner does not guarantee that the verification log is closed if the scanner is interrupted. 

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