[Mailman-Users] Slow Performance on semi-large lists

[J C Lawrence]

On Wed, 13 Dec 2000 15:34:30 -0700 (MST) 
D J Atkinson <dj at pcisys.net> wrote:

> If it's possible to do it with sendmail, desynchronizing DNS
> lookups would be one way.  [Back to Neil's question: Any body have
> any idea if/how to do this?] Another thought, that could help in
> my particular situation, is for qrunner to keep a record of what
> names have failed DNS lookup and only check those once during a
> given qrunner batch job.  This may not be feasible, depending on
> the software architecture.

There are some small points, such as you should be running a local
cacheing DNS server (I use DJBDNS, but BIND can do this as well).
After that, the point of diminishing returns starts becoming
apparent.  You can fix, patch, and tweak sendmail, but the question
of whether its really worth it any more becomes significant.

In the end you are going to move to a different MTA.  Depending on
the general expertise of your local SysAdms, I'd recommend Exim or
Postfix (Exim has some nice admin and monitoring tools that postfix
doesn't have yet, and has a very well documented and explained human
readableconfig file, Postfix is easier to setup and seems to perform
slightly better with default configurations, but requires more savvy
to tweak).

What to do about the posts sitting in your sendmail queue when you
do that?  Setup a second box to smarthost for your current sendmail
system.  Reconfigure sendmail on your first system to send all
outbound mail through that new box.  Let it run for a bit, and all
the outbound mail under sendmail should drain to the new smarthost.
Now install a new MTA on your list server, and let the smarthost run
until all its mail drains out.  You can then recommission the
smarthost.

-- 
J C Lawrence                                       claw at kanga.nu
---------(*)                        : http://www.kanga.nu/~claw/
--=| A man is as sane as he is dangerous to his environment |=--