Skip to contents

Get CPU and memory metrics for a SLURM job

script_metrics.Rd

Retrieves current resource usage statistics from SLURM commands including CPU utilization, memory consumption, and job status.

Usage

script_metrics(job)

Arguments

job

A parade_script_job object

Value

Named list with job metrics and resource usage

Details

This function queries SLURM's accounting system to retrieve job metrics:

Data Sources:

  • Uses sacct for historical/completed job metrics

  • Uses sstat for live metrics of running jobs

  • Falls back to squeue when accounting is unavailable

Metrics Returned:

  • cpu_pct: CPU utilization percentage

  • ave_rss, max_rss: Average/maximum resident set size (bytes)

  • ncpus/alloc_cpus: Number of allocated CPUs

  • elapsed: Elapsed time in seconds

  • node: Node list where job is running

  • state: Current job state (RUNNING, COMPLETED, FAILED, etc.)

Prerequisites:

  • SLURM commands (sacct, sstat, squeue) must be available in PATH

  • SLURM accounting must be enabled for detailed metrics

Failure Behavior:

  • Returns NA for unavailable metrics (never errors)

  • Warns once if SLURM commands are missing

  • Degrades gracefully when accounting is disabled

Examples

# \donttest{
if (Sys.which("squeue") != "") {
  job <- submit_slurm("script.R")
  metrics <- script_metrics(job)
  # Returns list with: cpu_pct, ave_rss, max_rss, elapsed, state, etc.
}
# }