[Condor-users] Proposal: Return a log of spawned processes

[Sfiligoi Igor <sfiligoi@xxxxxxxx>]

Dear Condor developers.

In my previous environment (the CDF experiment) we created a wrapper for 
our jobs that kept track of all the processes that were spawned from the 
main executable on the execute node, and logged that information in a 
log file that was finally transferred back to the submit node.

I know the Condor startd/starter is already tracking the spawning for 
other reasons, so I am wandering if it could also create such a log, if 
requested, and send it back to the submit node.
I would like to avoid creating another wrapper for this, especially 
since this seems to be completely generic.

Please let me know if you find this idea interesting.

If you do, I would just like to point out that you probably don't want 
to log very short lived processes. Experience tells, that a script can 
spawn thousands of awk, grep, ls and similar commands in very short 
periods of time, making the log file excessively large. If a user can 
specify the desired min duration (say 30seconds), the log file size is 
well under control.

Cheers,
 Igor