hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Luca Telloli (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-3133) [HOD] hodring.log-destination-uri should be set to a default value
Date Fri, 25 Apr 2008 17:19:56 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-3133?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12592434#action_12592434
] 

Luca Telloli commented on HADOOP-3133:
--------------------------------------

Additionally, log-destination-uri has a different behavior where set on hdfs:// or file://
. 
If used like hdfs://destination, the final tars will be under: destination/$USER/hod-logs/$JOBID

while on a file-system it would be just under destination. 

I think that the syntax for the [hodring] section should be enriched in such a way that  different
parameters/variables can be accessed for storing logs, like $USER or $JOBID, especially on
a non HDFS filesystem (where they'd take much less space). for instance by specifying a path
like file://dir/$USER/$JOBID. 

I made a test creating manually the dir/$USER directory, then passing the variable $USER with
log-destination-uri=file:///dir/$USER, and it worked, but since the job id is not visible
to bash, it is not possible to have a directory user/job. 

> [HOD] hodring.log-destination-uri should be set to a default value
> ------------------------------------------------------------------
>
>                 Key: HADOOP-3133
>                 URL: https://issues.apache.org/jira/browse/HADOOP-3133
>             Project: Hadoop Core
>          Issue Type: Improvement
>          Components: contrib/hod
>    Affects Versions: 0.16.0
>            Reporter: Hemanth Yamijala
>            Priority: Minor
>             Fix For: 0.18.0
>
>
> In the current setup, we do not insist a value for hodring.log-destination-uri which
is used to store tarred hadoop logs. Now, if there is a hadoop start up problem, the cluster
could get automatically deallocated. However, the hadoop logs would be lost too, as the above
configuration parameter does not have a default value.
> Possibly this should be set to atleast file:///<local file system> path so that
the logs will be saved.

-- 
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