[torqueusers] prevent log deletion / jobs with no time limit

Vincent LIARD vincent.liard at scilab.org
Fri Feb 19 08:25:09 MST 2010


Hello all,

There are two small problems I still haven't solved that some of you
could possibly answer easily. If you have some seconds to help me out...

1) How to preserve torque logs forever ?
(so that tracejob can be used on it any time)

As a first step, I tried:
# qmgr -c "set server log_file_roll_depth=30"
# qmgr -c 'p s' | grep log
set server log_events = 511
set server log_level = 7
set server log_file_roll_depth = 30

However, my log is still cleared every day... 
Shall I do something more ?

Moreover, how can I preserve logs indefinitely ? 
Would setting log_file_roll_depth to -1 do the trick ? 
(I seem to remember having read something like this but I don't manage
to find it back.)

2) How to allow my jobs to have no (cpu, wall) time limit by default ?

qsub documentation says "If not set for a generally available resource,
such as CPU time, the limit is infinite." 
However, when I set no limit tu cputime, there *is* a 1 hour limit. 
So, I use some high limit for these jobs (qsub -l cput=05:00:00).

For a more straightforward means, I tried cput=-1 butqsub frowns upon my
naivety (Illegal attribute or resource value for  Resource_List.cput)

I also tried cput=0. It looks plausible, but could you confirm it ? 
(I found nothing on this on
http://www.clusterresources.com/torquedocs21/2.1jobsubmission.shtml)

Anyway, how could I make it the default for all jobs ?

Many thanks in advance,

-- 
Vincent LIARD
Ingénieur de développement
--
Consortium Scilab
Digiteo
Domaine de Voluceau
Rocquencourt - B.P. 105
78153 Le Chesnay Cédex
Tél.: 01.39.63.55.47




More information about the torqueusers mailing list