On 08/11/16 19:23, Ran Du wrote:
> I am working on an accounting system to get a summary about how
> many cpus/gpus per node has been allocated to a specific job. However,
> all I can get from slurmdbd database is the total cpus/gpus allocated
> for a job, but not cpus/gpus allocated from each node. Does anyone know
> how to get these accounting information?
How is this?
$ sacct -j 6449483 -o jobid%20,jobname,alloccpus,allocgres
JobID JobName AllocCPUS AllocGRES
-------------------- ---------- ---------- ------------
6449483 wrap 6 mic:1
6449483.batch batch 6 mic:1
6449483.extern extern 6 mic:1
That's just a single node job asking for 6 cores and 1 MIC (KNC) card.
If you wanted more details you can use AllocTRES instead:
[samuel@barcoo Slurm]$ sacct -j 6449483 -o
jobid%20,jobname,alloctres%20,allocgres
JobID JobName AllocTRES AllocGRES
-------------------- ---------- -------------------- ------------
6449483 wrap cpu=6,mem=4G,node=1 mic:1
6449483.batch batch cpu=6,mem=4G,node=1 mic:1
6449483.extern extern cpu=6,mem=4G,node=1 mic:1
All the best,
Chris
--
Christopher Samuel Senior Systems Administrator
VLSCI - Victorian Life Sciences Computation Initiative
Email: [email protected] Phone: +61 (0)3 903 55545
http://www.vlsci.org.au/ http://twitter.com/vlsci
