[Mimedefang] Problem is happening right now (was: My MD install went wacko)

[Justin Shore]

On Wed, 11 Jun 2003, David F. Skoll wrote:

> Right.
> 
> You can easily check if the multiplexor is alive by doing (as root):
> 
> 	md-mx-ctrl status

I'll remove my restarting cronjob and wait for it to happen again so I can 
try this out.

> We have about 100 CanIt installations, and there are probably
> thousands of MIMEDefang installations, and I've never seen anyone
> reporting this lockup problem.  I suspect the Perl slaves are doing
> some kind of network operation like an RBL or Razor lookup -- if those
> take a long time, your slaves can get used up really quickly and
> everything grinds to a halt.

It is an odd one.  I've added numerous DNSBL checks to SA.  They use the
newer rbleval code added with 2.60 as well (at least that's what I've been
told since I can't find it documented anywhere :).  I may remove those BLs 
a handful at a time to see if they are the cause of the problems.  I 
haven't added any in a couple weeks though.  I can't think of any major 
change I made around the time this started happening.  I do recall seeing 
a message in the logs about no available slaves.  I didn't note the 
timestamp but I figured it corelated to when I bounce a thousand messages 
through my spamtrap and the load jumps to 30 or so.  That was my guess and 
I never investigated further.

> > Ah.  This would be beyond my skills.  I wonder if there's a debug mode for
> > the multiplexor...
> 
> Not one that's of use to end-users.  You can use the "-L" option to log
> slave status every few seconds; this will at least reassure you that
> the multiplexor is still sane.

The next time it gets in a rut and needs restarting every 5 minutes I'll 
add that option.  It might just turn up something useful.

I've tried downgrading SA to see if it was at fault.  That didn't seem to 
work.  I'll change my SA config back to a more basic config and see if 
that helps.

I'll also try lowering rbl_timeout.

After that I'll remove the Razor and Pyzor checks (no DCC at this time).

Along the way I'll liberally sprinkle in a few downgrades to SA 2.55 and 
upgrades to the latest greatest 2.60 to see if I get a working combo 
again.  I'll report back when I have some more information.  Thanks for 
the info

Justin