LinuxCommandLibrary

sstat

Display Slurm job or step statistics

TLDR

Display status information of a comma-separated list of jobs

$ sstat [[-j|--jobs]] [job_id]
copy

Display job ID, average CPU and average virtual memory size of a comma-separated list of jobs, with pipes as column delimiters
$ sstat [[-p|--parsable]] [[-j|--jobs]] [job_id] [[-o|--format]] [JobID,AveCPU,AveVMSize]
copy

Display list of fields available
$ sstat [[-e|--helpformat]]
copy

SYNOPSIS

sstat [options] job_id.step_id

PARAMETERS

-a, --all
    Display all fields.

-f, --format=
    Comma separated list of fields to display.

-i, --input=
    Read job step IDs from the specified file.

-j, --job=
    Specify the job step ID.

-n, --node=
    Specify the nodes to report statistics for.

-o, --output=
    Write output to the specified file.

-P, --parsable
    Display output in a parsable format.

-r, --raw
    Display raw values without human-readable formatting.

-s, --step=
    Specify the step ID to report statistics for.

-u, --user=
    Display statistics for specified user(s).

-v, --verbose
    Increase verbosity of output.

-w, --width=
    Specify the width of the output.

-H, --help
    Show help message and exit.

-V, --version
    Show version information and exit.

DESCRIPTION

The sstat command displays statistics about Slurm job steps. It retrieves information regarding resource usage, such as CPU time, memory consumption, and node utilization, for individual steps within a Slurm job. sstat is a powerful tool for analyzing job performance and identifying potential bottlenecks. Users can filter statistics based on job ID, step ID, node, and other criteria. This command is crucial for optimizing resource allocation and improving the efficiency of Slurm-managed clusters. The output can be customized to display various metrics and aggregated statistics. Analyzing step-level statistics is essential for understanding the performance characteristics of parallel applications and optimizing their resource requirements. With sstat users can gain insight into how their jobs are behaving on the cluster, allowing them to make informed decisions about resource allocation, code optimization, and job scheduling.

CAVEATS

The availability and accuracy of statistics depend on the Slurm accounting configuration. Data may be incomplete or unavailable if accounting is not properly configured.

EXAMPLES

Display statistics for job step 12345.0:
sstat 12345.0

Display CPU time and memory usage for job step 12345.0:
sstat -f cputime,maxrss 12345.0

SEE ALSO

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

Copied to clipboard