[thelist] Cold Fusion, Q-Mail, Spool Files

John Handelaar john at evolt.org.uk
Tue Feb 18 00:59:00 CST 2003


----- Original Message -----
From: "Frank" <lists at frankmarion.com>
To: <thelist at lists.evolt.org>
Sent: Tuesday, February 18, 2003 6:37 AM
Subject: Re: [thelist] Cold Fusion, Q-Mail, Spool Files


> I guess my next question would be this, then. Imagine that I'm looping
> though half a million rows in a database, each containing names, addresses
> and so on, merging it with text.
>
> Can I really call qmail-inject once for each row, pass it parameter and so
> on? Seems like asking it to do this once every 10 milliseconds or so is
> asking for a bit much.

Trying again.

If it really is a half-million addresses, you want to be using qmail-queue
to inject directly to the outbound spooler.  It requires that you add all
the headers yourself (whereas qmail-inject is a drop-in replacement
for /usr/bin/sendmail).

This guarantees (among other things) both that you don't lose cycles
on QMail trying to reverse-lookup the originating machine, and that
you're not slowed down by qmail-inject's processing of the messages.

At this volume, I would *strongly* recommend that you don't let this
Qmail server do the end-user delivery.  Chuck everything at a big fat
relay instead.

Base reference:  Qmail can famously (most quoted statistic) use a
Pentium-I class machine with 128MB of RAM and burn its way through
about 350,000 messages per day unassisted.  No doubt you have more
processor power and memory than that.  Bandwidth, however, may
be more of a speed issue for you than anything else in this case.

------------------------------------------
John Handelaar

T +44 20 8933 1494       M +44 7930 681789
F +44 870 169 7657   E john at userfrenzy.com
------------------------------------------






More information about the thelist mailing list