LinuxCommandLibrary

squeue

Display status of queued jobs

TLDR

View the queue

$ squeue
copy

View jobs queued by a specific user
$ squeue [[-u|--user]] [username]
copy

View the queue and refresh every 5 seconds
$ squeue [[-i|--iterate]] [5]
copy

View the queue with expected start times
$ squeue --start
copy

SYNOPSIS

squeue [options]

PARAMETERS

-a, --all
    Display information about all jobs, including those that have completed or are pending.

-u , --user=
    Display jobs belonging to the specified user(s).

-p , --partition=
    Display jobs running in the specified partition(s).

-j , --job=
    Display information about the specified job(s).

-t , --states=
    Display jobs in the specified state(s) (e.g., PENDING, RUNNING).

-o , --format=
    Specify the output format using field codes. Allows customization of displayed information.

-w , --nodelist=
    Display jobs running on the specified node(s).

-h, --noheader
    Do not print the header row.

--start
    Display the expected start time for pending jobs.

DESCRIPTION

The squeue command provides information about jobs in the SLURM job queue. It displays a snapshot of the current state of jobs, including job ID, user, job name, partition, state, submission time, and allocated resources. It is a crucial tool for monitoring job status, identifying potential bottlenecks, and managing workload in a SLURM-managed cluster.

squeue allows users to filter and sort the output based on various criteria such as user, partition, job state, and job ID. The command offers different output formats, allowing for easy integration with other tools and scripts. Understanding squeue is essential for efficient use of SLURM and for diagnosing issues related to job execution. It enables users and administrators to gain insights into the utilization of cluster resources and to optimize job scheduling strategies. Proper utilization can prevent resource starvation and ensure equitable access to cluster resources for all users. For a comprehensive understanding, consult the SLURM documentation.

CAVEATS

The output and behavior of squeue might vary slightly depending on the SLURM version and configuration. Some formatting options may not be available on all systems.

JOB STATES

Common job states include: PENDING (waiting to be scheduled), RUNNING (currently executing), COMPLETED (finished successfully), FAILED (terminated with an error), and CANCELLED (terminated by user or administrator).

OUTPUT FORMATTING

The '-o' or '--format' option offers extensive control over the displayed information. Use field codes (e.g., %j for JobID, %u for UserID, %T for State) to customize the output. Refer to the SLURM documentation for a complete list of field codes. For example: squeue -o 'JobID: %j User: %u State: %T'

HISTORY

squeue is a core component of the SLURM workload manager, developed to provide administrators and users with a way to monitor the status of jobs submitted to the cluster. It has been continuously improved and enhanced alongside SLURM to provide more detailed information and flexible filtering options. Its development has been driven by the need for better resource management and improved user experience in HPC environments.

SEE ALSO

sbatch(1), scancel(1), sinfo(1), scontrol(1)

Copied to clipboard