From: Ceylan, Tolga (Tolga.Ceylansavvis.net)
Date: Mon Feb 02 2004 - 12:44:29 CST

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

All;

I'm using postfix on 3 servers behind a load balancer. The problem is
that the load
balancer does not really check the server load, so as long as the server
accepts
TCP on port 25, it forwards SMTP traffic to it.

On several occasions one of the servers had a large queue, and was
unable to deliver the
mail as fast as it received. So, it's queue got larger, and larger.

For solve this problem I've written monitoring scripts to throttle the
incoming connections
by modifying master.cf, and limiting smtpd processes. Basically, the
scripts adjust
the smtpd process limit every 15/30 minutes, and reload postfix. In
case, the server gets a very
very big queue, then snmpd is disabled.This helps load balancer to
automatically take
the server out of the load balancing.

Everything worked perfect, however one of the servers were down after 2
days.
And I got this;

postfix/master[24511]: panic: master_spawn: at process limit 52

message... Now I understand that my solution would not work since master
freaks out when it sees a number of running snmpd processes more than my
limit in master.cf...

Is this a normal behaviour? Shouldn't master just not spawn new ones
until
the number of procs gets under the limit?

I'm running postfix-20010228release-3rh6.

If I upgrade to the latest, would this problem dissappear?

I appreciate any help on this...

Thank you,

Tolga Ceylan

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]