[torqueusers] qsig -s leaves processes running?
kenneth at sdsc.edu
Wed Oct 20 12:02:20 MDT 2010
I think this was with MVAPICH mpirun. Should it work
with OpenMPI mpiexec?
On Wed, 20 Oct 2010, Troy Baer wrote:
> Date: Wed, 20 Oct 2010 13:28:24 -0400
> From: Troy Baer <tbaer at utk.edu>
> Reply-To: Torque Users Mailing List <torqueusers at supercluster.org>
> To: Kenneth Yoshimoto <kenneth at sdsc.edu>,
> Torque Users Mailing List <torqueusers at supercluster.org>
> Subject: Re: [torqueusers] qsig -s leaves processes running?
> On Wed, 2010-10-20 at 09:27 -0700, Kenneth Yoshimoto wrote:
>> Sure, I was trying out qsig -s <job id> on a test system
>> last month with torque 2.5.2. The job went into S state,
>> but when I went out onto the compute nodes to check with
>> top, the processes of the job were still running.
> How were the processes spawned: directly by the job script, through some
> TM-aware program (e.g. pbsdsh, OSC mpiexec, OpenMPI mpiexec), or using
> an intermediary such as rsh or ssh? If it's the latter, TORQUE's
> pbs_mom has no knowledge of their existence and thus cannot deliver
> signals to them. OTOH, signal delivery to jobs and their associated
> processes does work in my experience, so long as all the processes are
> children of either the job script or pbs_mom.
More information about the torqueusers