Hi Oliver,
The source you have reserved are bit not match the apparently published
into the stream right ordered will be to put the SRCFG gap for the queue
class and USERCFG for *sgm, you may specify the HOSTLIST (FQDN) comma
separated to put in right order the othervise is up to you to build a
queue related policy and put them into maui.cfg
config sample:
NODESETPOLICY ONEOF
DESYNCTIME 0:00:10
DEACCESSPOLICY SHARED
JOBMAXOVERRUN 1
to give correctness deal for the clasess
USERWEIGHT 1
GROUPWEIGHT 1
........
USERCFG[<next*sgmFN>] HOSTLIST=FQDN1,FQDN2,...
USERCFG[<next*sgmFN>] PERIOD=INFINITY
USERCFG[<next*sgmFN>] TASKCOUNT=2 RESOURCES=PROCS:1
USERCFG[<next*sgmFN>] STARTTIME = 00:00:00 ENDTIME = 24:00:00
USERCFG[<next*sgmFN>] <rule for this/each specific user>
USERCFG[<next*sgmFN>] CLASSLIST=next class/(elsewhere queue name match)
Cheers
Paul
On Tue, 6 Feb 2007, Olivier van der Aa wrote:
> Dear All,
>
> At qmul we have the problem that standing reservation does not seem to
> work. They don't appear in showres -n
>
> When running diagnose -r at qmul we get this.
> "WARNING: reservation table is corrupt: active procs reserved does not
> equal active procs detected (1064 != 1448)"
> It seems that several jobs gets running on a node without the node being
> reserved in maui, for example:
>
> checkjob 763954
> -----
> checking job 763954
>
> State: Running
> Creds: user:cms012 group:cms class:lcg2_long qos:DEFAULT
> WallTime: 00:01:12 of 3:00:00:00
> SubmitTime: Tue Feb 6 07:33:20
> (Time Queued Total: 7:43:04 Eligible: 7:23:42)
>
> StartTime: Tue Feb 6 15:16:24
> Total Tasks: 1
>
> Req[0] TaskCount: 1 Partition: sl3
> Network: [NONE] Memory >= 0 Disk >= 0 Swap >= 0
> Opsys: [NONE] Arch: [NONE] Features: [sl3]
> Dedicated Resources Per Task: PROCS: 1 MEM: 500M
> Allocated Nodes:
> [cn235:1]
>
>
> IWD: [NONE] Executable: [NONE]
> Bypass: 0 StartCount: 201
> PartitionMask: [DEFAULT][sl3][installs]
> WARNING: active job has no reservation
> PE: 1.00 StartPriority: 443
> ----------------------------------------------------------
>
> Any idea of what can cause this ? It it like if maui was not in synch
> with what is actually happening on the cluster.
>
> Cheers, Olivier.
>
>
--
Dr. Paul A. Trepka ;Intl:+44(0)151 794 2137
Oliver Lodge Laboratory ;Fax: +44(0)151 794 3444
Dept. of Physics ;e-mail: [log in to unmask]
The University of Liverpool
Liverpool L69 7ZE
England, UK
|