Skip to content

Latest commit

 

History

History
55 lines (43 loc) · 1.82 KB

README.md

File metadata and controls

55 lines (43 loc) · 1.82 KB

Jobstats

Overview

Provides a toolkit of scripts to easily view and pipe data on a user or account's resource usage data from Slurm's sacct tool.

Usage

Usage:  jobstats [JOBID] [OPTIONS...]
Show usage statistics for past slurm jobs

  JOBID                a comma-separated list of jobids
  -A                   run jobstats on an entire account
  -S                   show jobs since start date (format m/d/yy)
  -r, --show-running   show running jobs
  -p, --parsable       make output parseable
  -c, --no-color       do not print colored stats
  -s, --no-size-limit  ignore dynamic terminal size limits
  -G, --gpu-stats      show gpu hours and gpu efficiency
  All other arguments in 'sacct' are accepted

Got bugs? Report to hpcsupport@nau.edu

Installation

Requirements

  1. Python 3.5+
  2. Slurm toolkit

Install

PyPi

pip install jobstats

Manual

git clone https://github.com/nauhpc/jobstats.git
cd jobstats
python3 setup.py install

Extremely manual

  1. Clone and enter repository git clone https://github.com/nauhpc/jobstats.git
  2. Add jobstats and group_efficiency to your PATH

Configuration

All config values for jobstats are accessed from jobstats-config.ini. These two files should always be in the same directory.

  • *_WEIGHT: Weight each value holds in calculating the average total efficiency.
  • SHOW_JOB_CHILDREN: Show each job along with the child jobs from each srun command in the batch script. Leaving this off will give you the sum total for each resource for the overall job.
  • COLUMN_MAX: Max width of the presented table. Setting to auto will set the max width to the terminal size on call.
  • GRADE_INCREMENT_*: Grade scale from red, to yellow, to green for each value.
  • TRACK_SINGLE_CORE_JOBS: Include single core jobs when calculating CPU efficiency