[Mauiusers] Requesting a given number of processors
Toni L. Harbaugh-Blackford [Contr]
harbaugh at ncifcrf.gov
Fri Sep 28 10:39:54 MDT 2007
On Fri, 28 Sep 2007, Jan Ploski wrote:
> mauiusers-bounces at supercluster.org schrieb am 09/28/2007 03:50:12 PM:
> > On Fri, 28 Sep 2007, Jan Ploski wrote:
> > > However, it still doesn't allocate correctly. I now tried with
> > > ncpus=9, which is more than the number of processors in any single
> > > node:
> > Ahh, as I said previously, you can't use ncpus as that's only for
> > requesting CPUs on a single CPU box.
> > You need to do nodes=9 to get what you need..
> I tried with nodes=9 now. A bit better than ncpus=9, but unfortunately
> still wrong:
I will look back at my configuration from a few months ago to figure out
how I did this. I actually had to *stop* my setup from doing the very thing
you want to do.
I know that in order to stop the behavior, I had to set nodect=1 and nodes=1 in
my qmgr setup. However, I don't know what else may be allowing this to work
when nodect=1 and nodes=1 are removed.
> Messages: cannot start job - RM failure, rc: 15044, msg: 'Resource
> temporarily unavailable REJHOST=node37 MSG=cannot allocate node 'node37'
> to job - node not currently available (nps needed/free: 4/1, joblist:
> PE: 9.00 StartPriority: -89
> job can run in partition DEFAULT (47 procs available. 9 procs required)
> For some unfathomable reason it is trying to allocate 4 procs on node37...
> Even though there are other nodes with 4 processors available... Or it
> could pick 1 processor from node37 and the rest from other nodes.
> Best regards,
> Jan Ploski
> mauiusers mailing list
> mauiusers at supercluster.org
Toni Harbaugh-Blackford harbaugh at ncifcrf.gov
Advanced Biomedical Computing Center (ABCC)
National Cancer Institute
Contractor - SAIC/Frederick
More information about the mauiusers