over 30 radiusd processes

[Alan DeKok]

Craig Campbell wrote:
> While the ps command doesn't show the time of these extra processes
> (over 24 hours old), in a previous event, I determined they seemed to
> coincide with a significant increase in radius traffic (from ~100/min to
> over 1000/min)  I believe the NAS forces all users to log off (and they
> automatically log back in) in the middle of the night, and I suspect
> this is related.

  Well, the server is *supposed* to clean up child processes.

> Alan suggested this might be related to shell scripts being run - as
> happens when acct STOP records are received.

  Yes.  If you can replace the shell script with a Perl script and use
rlm_perl, the issue will go away.

> I'm trying to come up with a strategy to narrow down what might be
> happening.

  The server isn't cleaning up child processes.  For some reason, it's
lost track of them.

> Under what circumstances does radiusd fork?  Also, I THOUGH I'd heard
> somewhere that threads and fork did NOT interact well.

  It forks when you exec a program.

  Alan DeKok.