Some of the Slurm commands can be a little clunky, especially if you start modifying their display output (using the -o
parameter that many of them support), so to help you out, we've created these additional aliases and shortcuts that are available systemwide for everyone to use.
Starts a remote shell (ie, an interactive job) on a compute node. This is equivalent to srun --pty bash
but still allows for providing additional resource parameters (queue, cpus, etc), which wouldn't be accepted by a normal alias in this specific use case:
$ srsh --partition=short
Lists the current state of the nodes, but with additional CPU allocation information (allocated/idle/other/total). This is equivalent to sinfo -o '%24N %.6D %.9P %.6t %.14C' -N
:
$ snodes NODELIST NODES PARTITION STATE CPUS(A/I/O/T) n19-04-008-cortana 1 debug idle 0/4/0/4 n19-32-192-ghost 1 short idle 0/64/0/64 n19-32-192-groot 1 short idle 0/64/0/64 ...
Provides historical post-job information (memory and CPU usage) for a job (with the -j <jobID>
) or all recent jobs if no arguments are given. This is equivalent to sacct --format=JobId,ReqMem,MaxRSS,AllocCPUS,TotalCPU,State,Elapsed --units=G
:
$ sjobacct -j 1000 JobID ReqMem MaxRSS AllocCPUS TotalCPU State Elapsed ------------ ---------- ---------- ---------- ---------- ---------- ---------- 630_369 3.91Gn 32 00:00.026 COMPLETED 00:01:01 630_369.bat+ 3.91Gn 0.00G 32 00:00.025 COMPLETED 00:01:01 630_369.ext+ 3.91Gn 0.00G 32 00:00:00 COMPLETED 00:01:01
Provides (limited) information on the maximum memory use of an active job. This is equivalent to sstat --format=JobId,MaxRSS
:
$ sjobstat -j 1000 JobID MaxRSS ------------ ---------- 637874.0 11580K