[Moabusers] Node access policies and reservations
wightman at clusterresources.com
wightman at clusterresources.com
Wed Jun 13 21:14:15 MDT 2007
We've tried this same setup and everything works correctly. Can you run the
command:
mschedctl -L 8
while a devhost job is sitting idle in the queue? Send us the output of that
command and we'll take a look.
Thanks,
- Douglas
On Tue, 12 Jun 2007, Michael Gutteridge wrote:
> I'm attempting to configure Moab (5.1.0p4) & Torque (2.1.8) to create a
> standing reservation for a few nodes which are intended for interactive
> development and testing work. These nodes will be shared, with multiple
> different people running interactive and non-interactive jobs.
>
> In torque, I'm configuring the nodes with "np=12" and setting up a queue
> called "devhost". Then in moab:
>
> CLASSCFG[devhost] QLIST=devhost QDEF=devhost
> CLASSCFG[devhost] MAX.WCLIMIT=6048000
> CLASSCFG[devhost] NODEACCESSPOLICY=SHARED
> QOSCFG[devhost] QFLAGS=USERESERVED:devhost
> SRCFG[devhost] ACCESS=DEDICATED CLASSLIST=devhost
> SRCFG[devhost] PERIOD=INFINITY
> SRCFG[devhost] HOSTLIST=zork22
> SRCFG[devhost] QOSLIST=devhost
> NODECFG[zork22] ACCESS=SHARED
>
> However, jobs submitted to the "devhost" queue don't run, just sit idle
> in the queue:
>
> Node Availability for Partition mercury --------
>
> zork22 rejected: Reserved (rsv devhost required)
> zork16 rejected: Reserved (rsv devhost required)
> NOTE: job cannot run in partition mercury (idle procs do not meet requirements : 0 of 1 procs found)
> idle procs: 13 feasible procs: 0
>
> I'd assumed that I'd configured the QOS or SR incorrectly, but this
> configuration works when I set "np=1" on the node.
>
> Has anyone run across this?
>
> Thanks
>
> Michael
>
>
> _______________________________________________
> moabusers mailing list
> moabusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/moabusers
>
More information about the moabusers
mailing list