[GE users] Dytran: Works from command line, not through SGE

Gerhard Venter gventer at sun.ac.za
Fri May 23 09:02:30 BST 2008


Reuti,

Thank you so much for all your input.  I finally got Dytran to run - I
actually feel pretty stupid since I should have figured this out a long
time ago.  Anyway, I finally looked at the command line options for
Dytran (as mentioned I am not a Dytran user myself) and found the
following option:

bat=no

which forces a non-interactive run.  So, when I modify my submit script
as follows everything works great:

#!/bin/bash
#$ -pe openmp 8
#$ -cwd
echo "Starting on" `hostname`
/apps/msc/dytran2007r1/bin/dytran bat=no jid=Klep72 ncpus=$NSLOTS

Again thanks for all the help and input and sorry for not picking this
up earlier.

Regards,
Gerhard

On Thu, 2008-05-22 at 17:41 +0200, Reuti wrote:
> Am 22.05.2008 um 15:48 schrieb Gerhard Venter:
> 
> > qrsh and qsub both result in the same error message.  However, when I
> > add strace -f to see where the problem is, the code runs!  Of course
> > strace then creates gigabytes worth of output.
> 
> Funny :-)
> 
> > The queue does not have any limits set that I am aware of.  The queue
> > definition is attached (obtained from qconf -sq default.q).
> 
> Maybe strace is forking the processes in a different way :-/
> 
> You can try to limit the output with "-e trace=file" or so. Maybe we  
> can get a clue, what is failing when we know the type of a call at  
> least.
> 
> -- Reuti
> 
> 
> > Gerhard
> >
> >
> > On Thu, 2008-05-22 at 14:41 +0200, Reuti wrote:
> >> Am 22.05.2008 um 13:34 schrieb Gerhard Venter:
> >>
> >>> Thanks for the feedback.  Unfortunately, I get the same error.  I  
> >>> few
> >>> other things I've tried in the mean time:
> >>>
> >>> 1.  Made sure I am mounting my work directory (on the head node)  
> >>> with
> >>> the sync option to NFS
> >>> 2.  Made sure the time is set consistent among all nodes
> >>> 3.  Tried running the script with ssh (something like "ssh
> >>> comp002 /home/blah/script") and it works without any issues
> >>> 4.  Tried qrsh and I get the same errors as with qsub
> >>
> >> As qrsh is working, can you try to run it via "strace" - maybe you
> >> can get some hint where it's crashing and why?
> >>
> >> Is there any limit in SGE on this particular queue (h_vmem...) set?
> >>
> >> -- Reuti
> >>
> >>
> >>> Many thanks,
> >>> Gerhard
> >>>
> >>> On Thu, 2008-05-22 at 10:24 +0200, Reuti wrote:
> >>>> Am 19.05.2008 um 18:43 schrieb Gerhard Venter:
> >>>>
> >>>>> I am very new to SGE and ran into an issue that I hope somebody
> >>>>> could
> >>>>> help me with - I hope I am posting on the correct mailing list.
> >>>>>
> >>>>> We are trying to run Dytran from MSC Software on our cluster.
> >>>>> When I
> >>>>> launch Dytran from the command line (on the head node or any of  
> >>>>> the
> >>>>> compute nodes) it works without any problems.  However, when I
> >>>>> launch
> >>>>> Dytran using SGE, I get an error message that states "Error  
> >>>>> closing
> >>>>> file".  We are using SGE 6.0 and OpenSuse 10.3 as the OS.
> >>>>>
> >>>>> My submit script is very simple, and looks something like this:
> >>>>> #!/bin/bash
> >>>>> #$ -cwd
> >>>>> /apps/msc/dytran2007r1/bin/dytran jid=Klep72
> >>>>>
> >>>>> As mentioned, I can run this script directly from the command line
> >>>>> without any issues.
> >>>>>
> >>>>> I am not an Dytran expert, but it seems that Dytran creates a csh
> >>>>> script
> >>>>> on the fly that it starts.  Could this be part of the problem?  I
> >>>>> cannot seem to find much help on using Dytran with SGE on the
> >>>>> web.  I
> >>>>> did try to explicitly set my limits in my submit script to be the
> >>>>> same
> >>>>> as from the command line (as mentioned in the FAQ on this website)
> >>>>> without any success.
> >>>>
> >>>> When this application creates a new job script on the fly, it might
> >>>> indeed jump out of SGE's control. But I never heard about such an
> >>>> error when closing a file. Maybe Dytran is using $TMPDIR which is
> >>>> removed after the job (but even then it's just like unlinking a  
> >>>> file
> >>>> and it should just disappear - unless they meaning "appending  
> >>>> some to
> >>>> the file to mark it as closed" or so.)
> >>>>
> >>>> To avoid the creation of an additonal shell (csh as kid of bash),
> >>>> maybe you can try the small adjustment and prepend an exec:
> >>>>
> >>>> exec /apps/msc/dytran2007r1/bin/dytran jid=Klep72
> >>>>
> >>>> This will replace the bash with your application, so one process  
> >>>> less
> >>>> in the process tree.
> >>>>
> >>>> -- Reuti
> >>>>> Any help will be greatly appreciated.
> >>>>>
> >>>>> Best regards,
> >>>>> Gerhard
> >>>>> -- 
> >>>>> +----------------------------------------------------------------- 
> >>>>> -+
> >>>>> || Prof. Gerhard Venter
> >>>>> ||
> >>>>> || Departement Meganiese en        |  Department of Mechanical and
> >>>>> ||   Megatroniese Ingenieurswese   |    Mechatronic Engineering
> >>>>> || Universiteit Stellenbosch       |  Stellenbosch  University
> >>>>> || Privaat Sak X1 Matieland 7602   |  Private Bag X1 Matieland  
> >>>>> 7602
> >>>>> || Suid-Afrika                     |  South Africa
> >>>>> ||
> >>>>> || Tel: +27 21 808 3560
> >>>>> || E-Mail: gventer at sun.ac.za          Web: www.eng.sun.ac.za
> >>>>> +----------------------------------------------------------------- 
> >>>>> -+
> >>>>>
> >>>>>
> >>>>> ------------------------------------------------------------------ 
> >>>>> --
> >>>>> -
> >>>>> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> >>>>> For additional commands, e-mail: users- 
> >>>>> help at gridengine.sunsource.net
> >>>>
> >>>>
> >>>> ------------------------------------------------------------------- 
> >>>> --
> >>>> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> >>>> For additional commands, e-mail: users- 
> >>>> help at gridengine.sunsource.net
> >>>>
> >>> -- 
> >>> +------------------------------------------------------------------+
> >>> || Prof. Gerhard Venter
> >>> ||
> >>> || Departement Meganiese en        |  Department of Mechanical and
> >>> ||   Megatroniese Ingenieurswese   |    Mechatronic Engineering
> >>> || Universiteit Stellenbosch       |  Stellenbosch  University
> >>> || Privaat Sak X1 Matieland 7602   |  Private Bag X1 Matieland 7602
> >>> || Suid-Afrika                     |  South Africa
> >>> ||
> >>> || Tel: +27 21 808 3560
> >>> || E-Mail: gventer at sun.ac.za          Web: www.eng.sun.ac.za
> >>> +------------------------------------------------------------------+
> >>>
> >>>
> >>> -------------------------------------------------------------------- 
> >>> -
> >>> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> >>> For additional commands, e-mail: users-help at gridengine.sunsource.net
> >>
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> >> For additional commands, e-mail: users-help at gridengine.sunsource.net
> >>
> > -- 
> > +------------------------------------------------------------------+
> > || Prof. Gerhard Venter
> > ||
> > || Departement Meganiese en        |  Department of Mechanical and
> > ||   Megatroniese Ingenieurswese   |    Mechatronic Engineering
> > || Universiteit Stellenbosch       |  Stellenbosch  University
> > || Privaat Sak X1 Matieland 7602   |  Private Bag X1 Matieland 7602
> > || Suid-Afrika                     |  South Africa
> > ||
> > || Tel: +27 21 808 3560
> > || E-Mail: gventer at sun.ac.za          Web: www.eng.sun.ac.za
> > +------------------------------------------------------------------ 
> > +<default.q>---------------------------------------------------------- 
> > -----------
> > To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> > For additional commands, e-mail: users-help at gridengine.sunsource.net
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net



---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list