An extension of the Slurm command 'seff' designed to handle job arrays and offers the option to display information in a histogram.
seff-array generates three types of histograms:
1. Maximum memory usage of each job in the array
2. Runtime of each job
3. CPU Utilization relative to the runtime
seff-array [-h] [-m] [-t] [-c] [-v] jobid
1. -m, --mem show memory usage stats
2. -t, --time show time usage stats
3. -c, --cpu show cpu usage stats
4. -v, --verbose enable to show histograms for job arrays
To use seff-array on the job array with ID 12345678, simply run seff-array 12345678.
The default when no flags given is -mtc (all stats).
Without the -v flag, seff-array produces output similar to the output of seff for both single jobs and job arrays.
Histograms and headings are designed to fit to 80% of your terminal window size.
Note: if your terminal window is too small (less than ~45 columns wide), then seff-array assumes
a terminal window of 100 columns wide.
Histogram bins vary according to the distribution of the data. There is a minimum bin width associated with each type of histogram to avoid unnecessary precision when the data is extremely clustered.
A series of simple tests will be tested with each push to this repository through CircleCI. The testing environment is composed of several key components:
- compare.sh - A basic bash script that calls seff-array and compares that output with the expected output.
- retrieve.py - Used in compare.sh. Returns the expected output of a given jobid.
- sacct - A mock sacct so that it returns raw sacct data (just like normal sacct does) given a jobid. The "fail" jobid is to simulate a scenario when sacct fails to return any output.
- tests/output - Directory that contains all expected output for each test.
- tests/data - Directory that contains all raw sacct data for each test.
- tests/log.txt - File that contains the mapping of each test to its jobid.