scontrol
scontrol is a command line tool used in high-performance computing clusters to manage and control the workload. It is a part of the Slurm Workload Manager, which is a highly scalable and flexible cluster management and job scheduling system.
With scontrol, administrators can configure and monitor various aspects of the cluster, such as scheduling policies, partitions, node properties, and accounting information. It allows users to submit and manage their jobs, view their progress, and control job resources.
The tool provides a wide range of options and subcommands to interact with the cluster. It allows users to set job priorities, specify resource requirements, change job attributes, cancel or hold jobs, and modify job dependencies. It also provides functionalities to manage node and partition configuration.
scontrol can retrieve detailed information about jobs, nodes, and other entities in the cluster. It supports filtering options to narrow down the output based on specific criteria. The tool also enables users to create scripts to automate job submission and management tasks.
By leveraging scontrol, administrators and users can efficiently control the workload in HPC clusters, ensuring optimal resource utilization and job performance.
List of commands for scontrol:
-
scontrol:tldr:1bd74 scontrol: Release a comma-separated list of suspended job.$ scontrol release ${job_id}try on your machineexplain this command
-
scontrol:tldr:7baed scontrol: Resume a comma-separated list of suspended jobs.$ scontrol resume ${job_id}try on your machineexplain this command
-
scontrol:tldr:a9a43 scontrol: Suspend a comma-separated list of running jobs.$ scontrol suspend ${job_id}try on your machineexplain this command
-
scontrol:tldr:c387d scontrol: Hold a comma-separated list of queued jobs (Use `release` command to permit the jobs to be scheduled).$ scontrol hold ${job_id}try on your machineexplain this command
-
scontrol:tldr:dc6a7 scontrol: Show information for job.$ scontrol show job ${job_id}try on your machineexplain this command