Len,

My first step was to set up WUG to 20 seconds. No improvements.
I know that maybe other MXs problems arise, but even by manually do postmap,
takes waaaaay long, around these 20 minutes I told.
And the maillog is *empty* from the minute 09:15:36 to the minute 09:34:01,
so even if the port 25 is listening, sounds like a nap for this while....
On monday, at 09:16 I'll do this 'sockstat -4' and let you know.

Regards and thanks

Andres

----- Mensaje original ----- 
De: "Len Conrad" <[EMAIL PROTECTED]>
Para: <[EMAIL PROTECTED]>
Enviado: s�bado, 15 de noviembre de 2003 13:10
Asunto: [IMGate] Re: Postmap hogs SMTP


>
>
> >The whole process takes about 20 minutes ( from the ftp to the postmap ),
> >but my monitoring system, What's Up, alert me each time the postmap
starts,
> >that the SMTP is not responding:
> >
> ># grep 09:15: /var/log/maillog
> >----snip-------
> >Nov 15 09:15:36 deimos postfix/smtpd[68639]: disconnect from
> >q5.elistas.net[216.66.19.153]
> >Nov 15 09:15:36 deimos postfix/smtpd[68616]: lookup table has changed --
> >exiting
> >Nov 15 09:15:36 deimos postfix/smtpd[68357]: lookup table has changed --
> >exiting
> >Nov 15 09:15:36 deimos postfix/smtpd[68613]: lookup table has changed --
> >exiting
> >Nov 15 09:15:36 deimos postfix/smtpd[68638]: lookup table has changed --
> >exiting
>
> those log line have nothing to do with postfix not listening. Each SMTPD
> process will terminate itself when it detects a .map.db change.  master.cf
> will start them up again.
>
> >----snip-------
> ># grep 09:33: /var/log/maillog
> ># grep 09:34: /var/log/maillog
> >Nov 15 09:34:01 deimos postfix/nqmgr[83093]: 65A6775E1B:
> >to=<[EMAIL PROTECTED]>, relay=none, delay=56824, status=deferred
(connect
> >to 127.0.1.50[127.0.1.50]: Can't assign requested address)
>
> 56834 is 16 hours and has nothing to do the postmap.
>
> # dig alians.com mx
>
> ; <<>> DiG 9.2.3rc3 <<>> alians.com mx
> ;; global options:  printcmd
> ;; Got answer:
> ;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 35986
> ;; flags: qr rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 4, ADDITIONAL: 0
>
> ;; QUESTION SECTION:
> ;alians.com.                    IN      MX
>
> ;; ANSWER SECTION:
> alians.com.             86400   IN      MX      0 127.0.1.50.
>
> their MX is screwed up. The rdata field is numeric, not a fqdn, and while
> postfix will try to handle (and log a warning for) a numeric MX, 127/8 is
> the loopback network, so NOBODY can deliver to alians' MX.
>
> drop this MX's problem from your analysis of the handling of the slet
file.
>
> >Nov 15 09:34:01 deimos postfix/nqmgr[83093]: 6D81875E97: from=<>,
size=3136,
> >nrcpt=1 (queue active)
> >
> >I can see it's not respondig. WUG it's right !!
>
> your WUG threshold is way too short for the checking IMGate's port
> 25.  make it 10 or 20 seconds to avoid false positives.
>
> >Any way to 'reduce' the 'postmap' impact ??
>
> when you postmap from_senders_slet.map manually, how long does it take?
>
> When an smtpd process starts up, it has to open all the .map files, do
> various consistency checks, and it can takes a few seconds. but it doesn't
> take minutes.
>
> so for a few seconds, postfix will be deaf, but not for minutes. but as
> soon as one smtpd process is reloaded, port 25 is alive.  during what you
> call the impact time (start it by do a postmap of the slet file), do this:
>
> sockstat -4
>
> to see if there any an smptd listening on *:25.
>
> and fix your WUG config.
>
> Len
>
>
>
>
>


Reply via email to