[torqueusers] Torque and Propack 5 cpusets.
csamuel at vpac.org
Thu Aug 9 22:54:29 MDT 2007
On Wed, 8 Aug 2007, Rushton Martin wrote:
> We run SLES 9 SP3, with PP4, SP3 on our Altix. I'm using cpusets
> in a simple manner. During boot an init script checks for the
> presence of the /batch cpuset, and if missing creates it. The
> three daemons (mom, server and sched) are then started within the
> cpuset, and will therefore create all children within it unless
> otherwise directed. We're (currently) using Torque only on a
> single 64-cpu node, not across a cluster so the procedure works
Hi Martin, nice to see Torque being used at the other fork of my old
Your method is handy for a large SSI system, but for clusters of
smaller SMP nodes it'd be nice if Torque would handle this - if for
no other reason than to catch the occasional rogue job that starts
heaps of processes accidentally.
At least if it were confined within its own cpuset it could only
hamper itself, rather than everyone else on the same node.
The hard part there would be handling processes that spread over nodes
via ssh/rsh rather than through the TM interface.
Be *really* nice to have a transparent rsh/ssh replacement that used
TM instead - could be a nice little project for someone (and no,
don't look at me!)...
I don't think pbsdsh is transparent enough, but that might be fixable
with a bit of hacking, perhaps switching into that mode if called as
Christopher Samuel - (03) 9925 4751 - Systems Manager
The Victorian Partnership for Advanced Computing
P.O. Box 201, Carlton South, VIC 3053, Australia
VPAC is a not-for-profit Registered Research Agency
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: This is a digitally signed message part.
Url : http://www.supercluster.org/pipermail/torqueusers/attachments/20070810/9bbc9033/attachment.bin
More information about the torqueusers