A very simple python script which monitors and records the CPU and RAM consumption of submitted cluster jobs.
To start recording use the
cpu_ram_log.py script. This script requires 3 arguments:
-uwhich corresponds to your username.
-owhich specifies the output file (in tsv format)
--intervalwhich specifies the time (in seconds) over which the CPU and RAM usage gets averaged
python cpu_ram_log.py -u nickhir -o cpu_ram.log --interval 5
The script can simply be included in your cluster job submittion like this:
#!/bin/bash #SBATCH --job-name=Example_run sambamba sort -t 20 some_bam.bam & MainScript=$! python cpu_ram_log.py -u nickhir -o cpu_ram.log --interval 5 & wait $MainScript pkill -P $$
Afterwards the resulting log file can be plotted (for that
matplotlib is needed).
For that use the
visualize_log.py script which needs 2 arguments:
-iwhich specifies the path to the log file which was created with
-owhich specifies the path to the resulting PDF which contains a plot for the CPU and RAM usage.
python visualize_log.py -i cpu_ram.log -o cpu_ram_visualization.pdf
The resulting plots will look somewhat like this: