[prev in list] [next in list] [prev in thread] [next in thread] 

List:       amanda-users
Subject:    Re: Run away processes?
From:       Tai <tai () sa ! fedex ! com>
Date:       2000-07-27 15:36:21
[Download RAW message or body]


"John R. Jackson" wrote:
> Actually, it **tries** (hard) to kill the dump process :-).  You wouldn't
> think this would be difficult, but some OS's make life, shall we say,
> interesting.

	Heh, I don't think it tried at all because all I have to do
as the operator is "kill $PID" and it goes bye bye.

> You should look at the /tmp/amanda/sendsize*debug and killpgrp*debug

	There is no killpgrp*.debug that matches the PIDs that stick
around.  All the killpgrp*'s from last night have this line:

killpgrp: debug 1 pid 26884 ruid 200 euid 200 start time Thu Jul 27
02:01:58 2000
/opt/amanda/libexec/killpgrp: version 2.4.1p1

	Well, the pid and time changed, but... :)

> files to track this further.  It could also just be a bug in vxdump that
> it's not killing off all its own children.

	This could be it.  Normally I wouldn't care, but in this case,
when the number of vxdumps that sit around get too high, it interferes,
so I guess I'll have top drop something like

% kill `ps -ef|grep vxdump|cut -c10-15`

	in a crontab.  I'm just not sure if I'm killing anything
that might not need killing.

> I run a script once a week to look around for all sorts of Amanda
> anomalies (e.g. file systems that have appeared and the admin forgot to

	My problem is that the PIDs of these beasties do not show
up anywhere.  Not on the machine doing the backup nor on the tape
box.

> slot.  They might be using some shared memory or semaphores, and that
> would eventually be a problem.  In fact, if you're killing them by hand,
> you should also run "ipcs -a" to see if anything is left around owned
> by the Amanda user that needs to be removed with "ipcrm".

	No, nothing is left around.  Thanx again for your kind help.
Guess I'll go with the crontab.  Oh well.

-Tai
["smime.p7s" (application/x-pkcs7-signature)]

[prev in list] [next in list] [prev in thread] [next in thread]