[Discuss] Corralling Processes on Linux

[kentborg at borg.org (Kent Borg)]

When I am dealing with files, I can easily delete them: if I am going to 
be creating cruft, I can put them all in a single directory, I can put 
subdirectories in that same directory, and when I am done, I can delete 
the whole thing, do it recursively, and they are gone. Easy.

Is there a way to do this with daemonized processes? I create an oddball 
collection, and want the ability to kill the whole lot.

Seems process group IDs might bark up this tree, but it doesn't look 
like I can tag a whole funny-shaped tree of processes with the same ID 
(this true?). And, my experiments along these lines have run into 
"operation not permitted"; I don't want to have to do this as root. (In 
the file example: I don't need to be root to put files and directories 
in a directory...)

I thought about dropping a unique string on the command line of each 
process. I won't be able to reliably find it with ps because it 
truncates, but I could crawl through /proc/*/cmdline and kill the ones I 
find. (Is there a way to do this on Mac OS?)

I thought about creating all these processes as a different user, and 
then killing everything owned by that user, but that probably requires 
root again (if that other user isn't me), and maybe I don't want to kill 
/everything/ (a login?) owned by the user.


Why I am doing this: I am playing with lots of different processes 
communicating with each other, maybe some coming and going 
incrementally. I want the ability occasionally kill them all and start 
from a clean slate. ("Oh, I had two of that one running for the last 
hour! Silly /me/ to waste all that time!" ...I want to avoid that set of 
bugs.)


Suggestions?


Thanks,

-kb