[Pacemaker] crm_mon SMTP notifications - message corruption

[Michael Schwartzkopff]

Am Sonntag, 15. November 2009 19:40:43 schrieb Lars Marowsky-Bree:
> On 2009-11-15T16:29:06, Michael Schwartzkopff <misch at multinet.de> wrote:
> > do you have the complete message header?
>
> The full message was attached, so you can see for yourself.

As far as I have learned SMTP messages headers include with a return-path, 
several receiveds, a date, a from a to and a Message-ID. At least.

For instance the full SMTP header of your first mail to the list started with:
Return-Path: <pacemaker-bounces at oss.clusterlabs.org>
Received: from mail.multinet.de ([unix socket])
	 by mail (Cyrus v2.2.12) with LMTPA;
	 Sun, 15 Nov 2009 14:51:10 +0100
X-Sieve: CMU Sieve 2.2
Received: from localhost (localhost [127.0.0.1])
	by mail.multinet.de (Postfix) with ESMTP id C4D061F659
	for <misch at multinet.de>; Sun, 15 Nov 2009 14:51:09 +0100 (CET)
Received: from mail.multinet.de ([127.0.0.1])
 by localhost (mail.multinet.de [127.0.0.1]) (amavisd-new, port 10024)
 with ESMTP id 09778-10 for <misch at multinet.de>;
 Sun, 15 Nov 2009 14:51:05 +0100 (CET)
Received: from oss.clusterlabs.org (oss.clusterlabs.org [75.127.97.36])
	(using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits))
	(Client did not present a certificate)
	by mail.multinet.de (Postfix) with ESMTP id 3434A1F62B
	for <misch at multinet.de>; Sun, 15 Nov 2009 14:51:05 +0100 (CET)
(...)

> > Can you count how many bytes the messages has including the messages
> > header? Perhaps something rips the message off after 1024 byte?
>
> This is not the case.

Perhaps yes, if you count all the header info.

> > Is it perhaps a problem with the broken SMTP installtion on the way (i.e.
> > Cisco Pix firewall)
>
> The recipient host is a postfix mail server.

Yes. And any Mail Interceptors or smarthosts on the way? I have seen 
especially Cisco PIX making considerable nonsense with SMTP connections. That 
was the basics of my question.

> > Did you trace the packets beeing sent out by the cluster with tcpdump?
> > Does this problem exists in the packages sent out from the nodes?
>
> I'd expect so. Where else would this occur?

On some device that "deepley inspects" the SMTP connection for security 
reasons. Not very likely if you have a direct connection, indeed.

> I'd need to recompile crm_mon with TLS disabled to make sense of the
> trace though.
>
> I was asking if other people had seen this before, so can I assume that
> SMTP alerts with crm_mon are working for you?

Sorry, I never tried SMTP. You perhaps will know that I am a admirer of SNMP.

Mit freundlichen Grüßen,

Michael.