sstat

TLDR

Display status of a running job

$ sstat -j [job_id]

Show specific fields in parsable format

$ sstat -p -j [job_id] -o JobID,AveCPU,AveVMSize

List available fields

$ sstat -e

Show all job steps

$ sstat -a -j [job_id]

Show PIDs for each step

$ sstat -i -j [job_id]

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.

PARAMETERS

-j, --jobs job[.step]

Job ID or job.step to query (required)

-o, --format, --fields fields

Comma-separated list of fields to display

-e, --helpformat

List available format fields

-a, --allsteps

Show all steps for specified job

-i, --pidformat

Display PIDs for each job step

-n, --noheader

Omit header from output

-p, --parsable

Pipe-delimited with trailing pipe

-P, --parsable2

Pipe-delimited without trailing pipe

--noconvert

Keep original units (don't convert)

-v, --verbose

Verbose output

-V, --version

Print version

COMMON FIELDS

JobID: Job identifierAveCPU: Average CPU timeAveRSS: Average resident set sizeAveVMSize: Average virtual memoryMaxRSS: Maximum RSSMaxVMSize: Maximum VM sizeNTasks: Number of tasksMinCPU: Minimum CPU time

INSTALL

sudo apt install slurm-client

CAVEATS

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.

HISTORY

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.

RESOURCES

Source code · Homepage · Documentation