sstat

sstat [OPTIONS...]

sstat displays real-time status information for running Slurm jobs including CPU usage, memory consumption, and I/O statistics. It queries the jobacct_gather plugin for metrics on currently executing jobs and their steps.
Root users can view all running jobs; regular users can only see their own jobs. The command is useful for monitoring resource usage during job execution and diagnosing performance issues.

-j, --jobs job[.step]

Job ID or job.step to query (required)

Comma-separated list of fields to display

List available format fields

Show all steps for specified job

Display PIDs for each job step

Omit header from output

Pipe-delimited with trailing pipe

Pipe-delimited without trailing pipe

Keep original units (don't convert)

Verbose output

Print version

JobID: Job identifier
AveCPU: Average CPU time
AveRSS: Average resident set size
AveVMSize: Average virtual memory
MaxRSS: Maximum RSS
MaxVMSize: Maximum VM size
NTasks: Number of tasks
MinCPU: Minimum CPU time

Only works on running jobs; use sacct for completed jobs. Requires the jobacct_gather plugin. Available metrics depend on the specific plugin (linux, cgroup). Non-root users have limited visibility.

sstat is part of Slurm, developed at Lawrence Livermore National Laboratory starting in 2002. Job accounting features help users and administrators monitor resource consumption in real-time. Slurm is now maintained by SchedMD.

sacct(1), squeue(1), scontrol(1), srun(1)