Hi, I am curious why no one has recommended using what looks like a built-in
testing [benchmark] tool in Postfix ?
/usr/sbin/smtp-source
I gave the man page a quick look and it seemed it may help, based on the given
specs provided by the OP.
I have versions 2.6.x and 2.10.x and assume it is still in versions 3.x.x
Good Luck in your testing.
-ANGELO FAZZINA
ang...@uconn.edu
University of Connecticut, ITS, SSG, Server Systems
860-486-9075
-----Original Message-----
From: owner-postfix-us...@postfix.org <owner-postfix-us...@postfix.org> On
Behalf Of Mauricio Tavares
Sent: Wednesday, March 6, 2019 9:20 AM
To: De Petter Mattheas <mattheas.depet...@jandenul.com>
Cc: Postfix users <postfix-users@postfix.org>
Subject: Re: stress tested postfix
On Wed, Mar 6, 2019 at 7:50 AM De Petter Mattheas
<mattheas.depet...@jandenul.com> wrote:
>
> Comments below
>
> Thanks for the help by the way :)
>
> -----Original Message-----
> From: patpro <pat...@patpro.net>
> Sent: 06 March 2019 13:34
> To: De Petter Mattheas <mattheas.depet...@jandenul.com>
> Cc: Postfix users <postfix-users@postfix.org>; owner-postfix-us...@postfix.org
> Subject: Re: stress tested postfix
>
> On 2019-03-06 13:10, De Petter Mattheas wrote:
>
>
> > Yes it is a strange business model, but the postfix must run on are
> > vessels.
> > So they sail over the world and because of the time difference we
> > can't help them ride away.
>
> I believe you should clarify or remake your test model, there is something
> I'm not sure to understand:
>
> - is a single ship likely to generate +40K email messages in an hour, or does
> this high figure applies to the postfix that will receive all messages from
> all ships?
> - do you account for the latency of offshore internet connection
> (satellite?) in your test?
>
> # a single ship can generate up to 60 K in a hour, we have had it in the past
> when a ship goes in error or failure or by mistake of the programmer it send
> alert mails to HQ
Meh, the Nigerian Prince can do better. ;)
> # the test was done in are virtual lab sow there was no SatCom involved, that
> was are next test. The interfaces where to virtual nic on a virtual switch.
>
>
> Also, to my experience high throughput email servers are heavily dependent on
> storage IO: you might find out that every other server performs OK on your
> virtualization node and still get an IO bottleneck in your email server.
>
> # still even when I/O is the bottleneck postfix should not accept the mail
> and leave it in que at the exchange, doesn't it ?
>
>
> Finally, like Wietse wrote, postfix does not lose email. Worse case
> scenario: your OS lose files. Normal scenario: postfix refuses email it can't
> ingest with temp error, sender retries later.
>
> # as I wrote before, than postfix should drop indeed the mail, and send
> retries.... but that doesn't happen. Postfix accepts the mail and exchange
> thinks its delivered. And the mail itself is lost not anymore on exchange
> and not in the maildir of postfix.
>
Could you crank up the postfix log to debug mode? I do remember
that when you let the log file be really verbose, it will keep track
of everything that happens to every single email. Then you have to
find which emails are going missing and follow them. Not a fun step
but has to be done.
If you can then find where in the sequence of events those emails are
disappearing, you might be able to figure out why. Or, posting that
part of the log will help others help you.
>
> Patrick
> ************************************************************
> Any reaction to this e-mail or any other mail, including any
> files transmitted therewith to sender's e-mail address(es)
> shall be dealt with not as private, but as business
> communication(s) and shall be registered as such.
> ************************************************************
>
