[torquedev] Are we ready to release a snapshot
garrick at usc.edu
Thu Jul 22 14:20:25 MDT 2010
On Thu, Jul 22, 2010 at 01:09:16PM -0700, Martin Siegert alleged:
> On Thu, Jul 22, 2010 at 01:32:38PM -0600, Ken Nielson wrote:
> > With the exception of the issue Stuart Barkley found I believe we have
> > all of our missing file issues and such taken care of. Are we ready to
> > release a snapshot (Not 2.5.2 yet) to be tested by everyone or does
> > someone have something else they want to get in?
> There is one feature I'd like to have:
> a) an environment variable that contains the number of processors
> assigned to the job. This appears to be easy to implement - patch
This comes up every now and then I always do my best to shoot down the idea of
implimenting this as an environmental variable.
One day we would like to have variable-sized jobs. If nodes/cpus are added or
removed while a job is running, then that env var can not be updated.
If implemented, it must be done with the count in a file and the name of the
file is in an env var. This is exactly like $PBS_NODEFILE.
> b) I'd like to be able to configure a routing queue where the jobs
> are routed to execution queues depending on the number of requested
> (total) processors, e.g.,
> np=1 -> q1
> 1 < np < 64 -> qs
> np >= 64 -> ql
> but now np is not a requested resources anymore but the sum of
> the nodes and procs request, e.g., in the example above np=10.
If "procs" gets traction, then I'm sure this can be done with procs.
Garrick Staples, GNU/Linux HPCC SysAdmin
University of Southern California
Life is Good!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: not available
Url : http://www.supercluster.org/pipermail/torquedev/attachments/20100722/bce21196/attachment.bin
More information about the torquedev