Draft:Jobhtop Overview
For more information, check out https://man7.org/linux/man-pages/man1/htop.1.html and https://peteris.rocks/blog/htop/#mem-memory-usage)
The command jobhtop <jobID> gives more details about your job. Specifically, it shows the resource usage of your job so you can determine if your job needs more/less resources and can make the necessary changes.
Once in the jobhtop screen, pressing F2 will allow you to customize which columns and information you want displayed about your job.
- Press F2 for Setup
- Under the green Setup column, use the down key to go to Columns, press Enter
- There, use the arrow keys to move over to Available Columns, and press F5 (Add) to add which information you want displayed on the main jobhtop page
- Press F10 once you’re done customizing
Meaning of Columns:
- MEM% - Shows how much memory a job is using out of available physical memory
- CPU% - Percentage of the CPU time the process is currently using; 100% means it is using the full capacity of the core
- I/O Rate - Shows the rate of data transfer in bytes per second
Using jobhtop is different if you want to check on a job array. You need to first ssh into the node that the job is running on. The command squeuemine will show a list of your jobs and what compute node they are running on.
[user@login2 ~] $ squeuemine
JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON)
87642467 c0706a-s5
[user@login2 ~] $ ssh c0706a-s5
[user@c0706a-s5 ~] $
Now you can type htop, type u, and select your username and check on your job. When you are done, hit q to quit and then type exit to logout. You should not run anything other than checking on your job while logged into a compute node.