|
Written on: 03. 02. 2012 [12:06]
|
|
hhoffmann
Holger Hoffmann
Topic creator
registered since: 21.12.2011
Posts: 30
|
Hi,
I'm having this issue:
I have calculations on time series (climate data) for n sites and although the data has always the same structure, the duration of jobs is very different. So depending on 1.) The host machine 2.) The data, most of the jobs (~70%) complete in 6 h, but the rest may last up to 20 h.
Therefore I get these complain e-mails to adjust the requested wallclocktime. I cant solve it, because it would be very exhausting to make pre-runs for every site, nor do I know the machine it will run on.
Is this an issue or should I just delete these mails everyday? How do you handle this?
Kind regards,
Holger
|
|
Written on: 03. 02. 2012 [12:25]
|
|
gerdes
Andreas Gerdes
registered since: 14.09.2010
Posts: 50
|
Hi Holger,
in your case: just ignore the e-mails.
Best regards
Andreas
|
|
Written on: 21. 02. 2012 [19:52]
|
|
hhoffmann
Holger Hoffmann
Topic creator
registered since: 21.12.2011
Posts: 30
|
Hi,
second wallclocktime-problem: I´m getting jobs aborted. E.g.:
Job: reRun3.o473621-492
--------------------------------------------------------------
Job ran on: lucky.rrzn.uni-hannover.de
Module for Mathworks Matlab, version 2011b loaded.
=>> PBS: job killed: walltime 144036 exceeded limit 144000
--------------------------------------------------------------
A 40 h - Wallclocktime-job, supposely.
The problem: It ran only <20 h, according to the shell:
qstat -t 473621[]
How come?
Kind regards,
Holger
[This article was edited 1 times, at last 21.02.2012 at 19:53.]
|
Top
Print