[
https://issues.apache.org/jira/browse/HDFS-348?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13066719#comment-13066719
]
Harsh J commented on HDFS-348:
------------------------------
I disagree that the NN be responsible for a DN's fault and be aware and log
those. Its additional work, given that the exact failed machine address is
bubbled to the client with the error to debug further.
Why must the NN log this when it isn't directly the one responsible (all it
does is hand over locations, and that's how independent it ought to be in all
situations)?
> When a HDFS client fails to read a block (due to server failure) the namenode
> should log this
> ---------------------------------------------------------------------------------------------
>
> Key: HDFS-348
> URL: https://issues.apache.org/jira/browse/HDFS-348
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: eric baldeschwieler
> Assignee: Sameer Paranjpye
>
> Right now only client debugging info is available. The fact that the client
> node needed to execute a failure mitigation strategy should be logged
> centrally so we can do analysis.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
