Have a process called 'software' that sometimes has an issue with an image, here called 'image.png'

When this happens, 'software' will die, and restart, over and over and over, constantly changing PIDs - but keeping the same image name 'image.png'

What would be your suggestiion for monitoring for this case?

find the binary executable file and chmod it to non-executable.

1) That isn't monitoring / testing for a condition
2) That will keep it from working at all

I do not know what is your goal, what do you want to monitor and why. But I would look for its parent process first.

1 members found this post helpful.

I explained it up there in the first post. An example would be this:
See that process and filename? A second later:

See that? The pid changed, but 'software' is trying to work on image.png again?

Keeping in mind that image.png might be anything(image01.png or thisimage.png perhaps) - What is your suggested way of detecting this failure?

Does it generate any messages to a logfile when it "fails" and restarts?

You do not say what process invokes software originally, whether it works through some directory of images, is it cron driven or attended, etc.

Suggestions might be to tail messages and look for repetitions, or modify the software to create a lockfile based on or including the image filename which would block or abort, or maybe write a wrapper script to detect and abort.

If PID (process info) is the ONLY way of identifying it, then write a script to ps | grep | compare last ? fail : continue | save last

1 members found this post helpful.

First of all, you should find out who (which user) runs this program, what does it do (or is it supposed to do), is this behaviour considered to be ok?

1 members found this post helpful.

Will do. Thanks for the hints.

Parent and child share buffers and file descriptors. Maybe /tmp/image.jpg belongs to parent process forking in reincarnation of "software". I would try lsof for /tmp/image.pgn.

1 members found this post helpful.

sometimes you can remove that image.png and the "software" will drop an error message - and that will help you to identify it.

1 members found this post helpful.

Wait, is the actual name of the binary program "software"? If so, simply remove the program-file, remove the user that owns it, and remove every other files of that user. It must be some idiot having fun with your computer.

1 members found this post helpful.

How are you running this? Did you write "software".

Actually, forget that, you can write a program to run "software" from within a C program and check for signals. See this blog about a daemon and how to have it monitor the children. Start down at the chunk about creating child processes and monitoring them. Check exit status, and so forth.

And if something dies, it shouldn't be recreating itself, there's a controller, daemon, or service which is already monitoring it and restarting it.

1 members found this post helpful.