[torquedev] pbsdsh and number of processors allocated

Michael Barnes barnes at jlab.org
Thu Jun 18 08:48:47 MDT 2009


On Thu, Jun 18, 2009 at 12:05:56PM +0100, Craig Macdonald wrote:
> Would there be some mechanism for notifying running processes of a 
> change? Or would they be expected to poll a file. Lets say it's called 
> $PBS_CPUCOUNTFILE, and just contains the number of processors that the 
> job has allocated.

Ever since FUSE came around I always thought it would be nice if there
were a PBS_FS FUSE module.  This would be simple to implement for modern
OSes like Linux, OSX, and FreeBSD.  I would see it like $PBS_O_WORKDIR
to have directories like:

JOBID

and inside of the JOBID there would be things like /proc/cpuinfo for the
CPU, things like:

state: queued
cpus: 8
nodes: 2
qtime: 12345678
mtime: 12345678
ctime: 12345678
cpuset: /Charlie
mom_host: node01.here
this_host: node02.here
hostlist: node01.here node02.here

or whatever people feel as necessary.  People are already using awk,
perl, sed, and cat to get some of this info from qstat, pbsnodes, the
$PBS_NODEFILE, etc.  All of these things are hacks.  I would believe it
would be nice to have a more unified /proc style interface to the batch
system.

Environment variables are a PITA.  They don't survive going across
machines very well, and they are difficult to get the right environment
variable on this host vs that host.

I've been looking at FUSE for some time, and its only had fairly
pedagogical/proof of concept uses thus far, and I've always felt that
something like a batch system could take advantage of an overlayed FUSE
filesystem make things more coherant.

-mb

-- 
+-----------------------------------------------
| Michael Barnes
|
| Thomas Jefferson National Accelerator Facility
| 12000 Jefferson Ave.
| Newport News, VA 23606
| (757) 269-7634
+-----------------------------------------------


More information about the torquedev mailing list