[torquedev] possible bug in qstat -f

Pablo Fernandez pablo.fernandez at cscs.ch
Fri Sep 16 09:35:38 MDT 2011

Dear devels,

We have been fighting with a third-party script that depends on the output of 
qstat -f... and it seems the output format has changed slightly, and the 
script does not work anymore. We have changed the script to make it work, but 
as I mentioned, is 3-party, so no further updates are possible.

Besides, it really seems to me like a typo somewhere in the pbs client code 
(2.4.16), so I thought it may be worth sharing it here.

So, if you type qstat -f and you show all special characters, you get this:

Job Id: 5304000.lrms02.lcg.cscs.ch$
   Job_Name = cre01_914471456$
   Job_Owner = lhcbprd01 at cream01.lcg.cscs.ch$
   job_state = Q$
   queue = lhcb$
   server = lrms02.lcg.cscs.ch$
   Checkpoint = u$
   ctime = Sat Sep 10 00:10:41 2011$
   Error_Path = cream01.lcg.cscs.ch:/dev/null$
   Output_Path = cream01.lcg.cscs.ch:/dev/null$
   Priority = 0$
   euser = lhcbprd01$
   egroup = lhcb$
   hashname = 5304000.lrms02.lcg.cscs.ch$
   queue_rank = 1300167$
   queue_type = E$
   comment = job rejected by RM 'lrms02' - job started on hostlist wn176.lcg.$
^Icscs.ch at time 15:20:19_09/16,$
^I job reported idle at time 15:23:14_09/16 (see RM logs for details)$
   etime = Fri Sep 16 15:50:32 2011$
   submit_args = /tmp/cre01_914471456$
   fault_tolerant = False$
Job Id: 5304708.lrms02.lcg.cscs.ch$
(the next job continues...)

So, if you take a look closely, the "comment=" line finishes with two returns, 
where all the other fields finish with one. So, this means any parser that 
expects a blank line to separate two jobs, will fail.

I guess the comment line was introduced in the 2.4 series, because the scripts 
were designed for 2.3, and they don't fail there.

Is this actually a bug that could be fixed within the same series (2.4)?

Thanks a lot!

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torquedev/attachments/20110916/d52e9527/attachment.html 

More information about the torquedev mailing list