Parse sinfo output to map GPU resources across Slurm cluster partitions.
npx skills add https://github.com/TzJ2006/gadget --skill slurm-gpu
Please help me install this Agent Skill. Skill: slurm-gpu Source: https://github.com/TzJ2006/gadget/tree/main/skills/slurm-gpu Command: npx skills add https://github.com/TzJ2006/gadget --skill slurm-gpu
Identify and map GPU resources across a Slurm cluster by parsing sinfo output and current allocations.
Run /slurm-gpu to view current GPU availability and partitions across the cluster.
High-intent search queries and answers about installing and using this skill.
To check available GPU resources across a Slurm cluster, you can parse sinfo output and current allocations. This identifies idle, allocated, mixed, and down states to produce a structured report of per-node and per-partition GPU details.
Slurm cluster monitoring determines GPU states by parsing sinfo output alongside current squeue allocations. It evaluates nodes to distinguish between idle, alloc, mix, and down states, returning detailed per-node GPU availability.
Yes, you can filter Slurm GPU outputs to show free GPUs in specific partitions. Applying filters like --free or --partition refines the structured report, directly supporting job submission decisions and resource planning.
No, this GPU cluster monitoring approach does not require write access to the Slurm scheduler. The tool operates in read-only mode, validates Slurm availability, and safely parses sinfo and squeue outputs without modifying resource allocations.
The best way to inventory GPUs by type and partition using Slurm is parsing sinfo output and current squeue allocations. This approach maps GPU resources across the cluster, detailing idle and allocated states per node.
Slurm GPU reports show mixed states instead of fully idle resources when a node has some GPUs allocated while others remain free. The tool handles idle, alloc, mix, and down states to accurately reflect current cluster allocations.
carbonscottAnswer Slurm questions on SLAC's S3DF cluster with read-only commands.
jimmc414Run Python code in serverless cloud containers with GPUs and autoscaling.
umich-foreseerDiagnose Slurm job failures by analyzing sacct, scontrol, and logs.
cnoe-ioQuery EKS clusters via kubectl and report pod, node, and resource health.
