Postfix/Monitoring_postfix_processes

From http://archives.neohapsis.com/archives/postfix/2008-10/0602.html

User wants to make sure Postfix is running, probably like what we do at work for NRPE process monitoring.

I reckon monitoring mail arrival would be fine. NRPE can fire sendmail with a timestamp in the subject (or maybe just rely on header timestamps), and a passive check can calculate the arrival delay. The check would probably cause an alert for freshness expiration too.

> We run Postfix on most servers and would like to have accurate monitoring of
> the various processes that are running.

I take a different approach, I send probe messages every ~30s and monitor
their arrival time. If everything is running that needs to run, messages
arrive. Otherwise, they stop.

> The monitoring has configuration for processes that "must" be running (and
> will alert if they are not) and those "allowed" to be running (and will
> alert if anything not present in either list is running).

The only required process is "master". Everything else is restarted on
demand, but in the queue manager, such restarts must not be too frequent.

> I've read http://www.postfix.org/OVERVIEW.html and can compile a list of
> Postfix components but my question relates to which ones should be running
> 100% of the time and therefore should be in our "must" configuration.

Just "master".

> My understanding is that only 'master', 'smtpd' and 'qmgr' fit this criteria
> but any clarification the list can offer would be appreciated.

When no clients are connecting, after $max_idle, there will be no smtpd(8)
processes. A minimal Postfix installation will have just:

     PID    PPID  PGID USER     COMMAND
     11549     1 11549 root     master
     11551 11549 11549 postfix  qmgr
      6402 11549 11549 postfix  pickup

with all other daemons launched on demand, but it is really sufficient
to make sure mail is being delivered (and perhaps that master is
running). Don't worry about individual processes.