Page 1 of 1

PX job re-run abort

Posted: Fri Jan 23, 2004 2:37 am
by praj
I am running a PX job creating multiple datasets from a seq file using a Xmer. The input has millions of records. Now when i run this job for the first time, there was no problem but when i rerun that job changing input and without recompiling job it aborted.
if i compile the job, it works. The error it gives is "Sort_101,0: Fork failed: Not enough space"

ne help
cheers
praj

Posted: Fri Jan 23, 2004 4:28 pm
by ray.wurlod
Others have encountered this, as a search of the Forum will reveal. It's probably a lack of scratch space on one or more of your nodes. The solution is to allocate more and/or to clean out unneeded files from any scratch space that you've already allocated.

Re: PX job re-run abort

Posted: Fri Jan 23, 2004 4:29 pm
by Teej
praj wrote:I am running a PX job creating multiple datasets from a seq file using a Xmer. The input has millions of records. Now when i run this job for the first time, there was no problem but when i rerun that job changing input and without recompiling job it aborted.
if i compile the job, it works. The error it gives is "Sort_101,0: Fork failed: Not enough space"
What the error you provided is saying: When you are attempting to run a Sort on whatever data you have, it apparently ran out of scratch space.

Double check to ensure your scratch space is big enough for the data. You CAN use multiple mountpoint per node within your configuration file. Do a search on this forum for configuration files.

-T.J.

Re: PX job re-run abort

Posted: Fri Jan 23, 2004 11:31 pm
by vzoubov
praj wrote:Iit gives is "Sort_101,0: Fork failed: Not enough space"
Not a Unix expert but sounds to me like there's not enough virtual memory to start a new process.

Vitali.