[torquedev] nodes, procs, tpn and ncpus
glen.beane at gmail.com
Mon Jun 14 19:33:37 MDT 2010
On Mon, Jun 14, 2010 at 9:10 PM, David Singleton
<David.Singleton at anu.edu.au> wrote:
> On 06/15/2010 09:42 AM, Christopher Samuel wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>> On 11/06/10 19:01, "Mgr. ï¿½imon Tï¿½th" wrote:
>>> What user? Unless he is using Maui/Moab then he means
>>> four nodes. And if the user is using Maui/Moab, then
>>> this user is irrelevant to this discussion (cause all
>>> external schedulers send very specific run requests,
>>> that shadow all of the original Torque semantics).
>> Oh, I didn't realise anyone used pbs_sched these days, sorry!
>> Point taken.
> I think some of the discussion is coloured by even more basic
> functionality than that - in particular, what happens when an admin
> runs qrun. Does anyone actually do that? I think PBS Pro stopped
> this a while back. And I'd imagine it would not help most schedulers
> to have random jobs starting under them (they should still handle it
> but it may not be ideal). If we were to support qrun then it would
> probably just crank up the jobs priority and kick the scheduler.
personally I think qrun is helpful for TORQUE developers, but not the
best thing to do on a production cluster
More information about the torquedev