-Re: issue with hadoop mapreduce using same job.jar
Ben Kim 2013-03-12, 16:22
Things I tried so far without a luck
- restart the hadoop
- sync && echo 3 > /proc/sys/vm/drop_caches
- clear namenode java cache using jcontrol
- check permission of the /user/hadoop/.staging folder
- delete everything under the .staging folder
- rename the test.jar file
- run using different user
what worked though is remotely running a MR using the Hadoop API
so it seems like this is only happenening in the Hadoop CLI
On Wed, Mar 13, 2013 at 1:00 AM, Ben Kim <[EMAIL PROTECTED]> wrote:
> Hi there
> It looks like the job.jar created in the /user/hadoop/.staging/ folder is
> always the same no matter which jar file i give.
> if i download the job.jar file then I reckon that it's a jar file I used
> run a MR job few hours ago.
> I'm using hadoop 1.0.3 on top of centos 6.2
> anyone has any ideas?
> *Benjamin Kim*
> *benkimkimben at gmail*
*benkimkimben at gmail*