[slurm-users] How to request for the allocation of scratch .

[navin srivastava]

Hi Team,

i wanted to define a mechanism to request the local disk space while
submitting the job.

we have dedicated /scratch of 1.2 TB file system for the execution of the
job on each of the compute nodes other than / and other file system.
i have defined in slurm.conf as TmpFS=/scratch  and then wanted to use
#SBATCH --scratch =10GB   in the request.
but it seems it is not accepting this variable except /tmp.

Then i have opted the mechanism of gres.conf

GresTypes=gpu,scratch

and defined each node the scratch value and then requested using
--gres=lscratch:10GB
but in this scenario if requesting both gres resources gpu as well as
scratch it show me only scratch in my Gres resource not gpu.
does it using the gpu also as a gres resource?

could anybody please advice which is the correct method to achieve the same?
Also, is scratch will be able to calculate the actual usage value on the
node.

REgards
Navin.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20200413/5638ab02/attachment.htm>