[
https://issues.apache.org/jira/browse/HADOOP-3239?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12588130#action_12588130
]
Doug Cutting commented on HADOOP-3239:
--------------------------------------
This issue is about log messages generated by calls to exists() that didn't
used to be generated by calls to exists(). It is a regression. It is not
about namenode exception logging more generally. (About that, however, I think
folks have found it convenient that whenever an exception is received from RPC
that it is also logged on the server side.)
> getFileInfo() used to return null.
That's not true. It has always thrown an exception for this case, since that
was always before the contract of its corresponding FileSystem methods. Now
that it is overloaded to support exists() too, it should no longer throw the
exception at this level.
http://svn.apache.org/viewvc/lucene/hadoop/trunk/src/java/org/apache/hadoop/dfs/FSDirectory.java?r1=549977&r2=549976&pathrev=549977
> exists() calls logs FileNotFoundException in namenode log
> ---------------------------------------------------------
>
> Key: HADOOP-3239
> URL: https://issues.apache.org/jira/browse/HADOOP-3239
> Project: Hadoop Core
> Issue Type: Bug
> Affects Versions: 0.17.0
> Reporter: lohit vijayarenu
>
> exists() was modified to invoke getFileStatus() internally. But
> getFileStatus() throws FileNotFoundException for files which does not exists
> and this is logged in RPC$Server for each exists() call. One way to get rid
> of these messages is at the Name Node, catch FileNotFoundException and return
> null. In this case, RPC would not log it in namenode log. But at the client
> end we might have to check for null on all calls of getFileStatus(). Other
> option at client end is to construct FileNotFoundException() when
> getFileInfo() returns null.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
