multicore-sys-monitor Feature Request: Add an option to show warning if a process is HANG (nearly 100% load on the same core for a long time)

[Heavygrass]

Applet name and maintainer

multicore-sys-monitor@ccadeptic23 @claudiux

What would you like to see?

On multi-core systems it is hard to notice if a particular process is using (nearly) 100% load on the same core for a long time (with average load graph).
On multi-core systems overall average load still remains pretty low even when there's one hang process, and I often don't realize that there's a hang process for a day or more.

One could determine there's a hang process if it loads the SAME core for say 10 minutes or more (exact threshold period is subjective and preferably be a field to specify).
Short time 100% load on the same core is normal of course.

This does NOT ask to monitor every single process, just the load on cores.

This is especially important for laptops.
Thank you.

[claudiux]

If I use the result of cat /proc/stat | grep procs_blocked, is it OK for you?

[Heavygrass]

@claudiux
Thank you for checking with me!

linuxhowtos says "The "procs_blocked" line gives the number of processes currently blocked, waiting for I/O to complete."
A process "waiting for I/O to complete." shouldn't load CPU much (if at all), so that's not what I'm looking for (and probably won't be interesting for other people).

I'm not sure if there's a way to detect long-term high same core load besides calculating an average load over time (separately for each core).
1 sample say per 10 sec should be enough.
There might be a way to do this by only keeping about 3 values in memory per core (in case there's a concern of samples taking RAM, don't really see it as an issue):

• current average
• only very first sample for the current interval
• only most recent in time sample

One could just update the average with very few calculations at each timestep (separately for each core).

[claudiux]

@Heavygrass
Thank you; now I know what to do. Please be patient, I'm busy this time.
I plan to change the CPUs box background color (set by user) when a process is hang.

[Heavygrass]

@claudiux
Just a suggestion: a common simple data structure like linked-list might be simpler to use for storing samples comparing to array.
It can also help to avoid unnecessary moving/copying of data when adding a new sample or deleting a no longer relevant old sample.

[claudiux]

Any suggestion or part of code is welcome.