Hi,
First, a bit of background:
I have a Debian 2.2 box running Sendmail 8.11.3+8.12.0.Beta7-3
There are a bit under 400 domains that this box accepts mail for and it's
using virtusertable to map addresses in these 400 domains to addresses in
another domain (that isn't hosted on the box)
So the general flow of mail is that anything addressed to one of these 400
domains lands on this box, gets translated and then onsent to the single
domain (lets call it xyz.com).
This is where I'm running into all sorts of problems. I'm getting hundreds
of messages queued up on my mailserver that are having problems getting
delivered to the MX for xyz.com. The three main error types are:
(I/O error: Input/output error)
(Deferred: Connection reset by mx.xyz.com)
and something about broken pipes
If I go and run the queue manually (sendmail -v -q) it will usually
deliver the first message destined for mx.xyz.com and then spray the above
type error messages for the rest of them.
If, however, I run the queue individually[1], most of the messages in the
queue get delivered without a problem.
So I've had to resort to "running the queue" by putting in a cron job that
runs every individual message in the queue rather than letting sendmail do
it's builtin thing automatically every 10 minutes.
I just found the whole thing odd, and after spending a fair bit of time
poking sendmail and increasing loglevels and not really getting anywhere,
I'd ask around for suggestions.
Andrew
[1] by running this:
for i in $(cd /var/spool/mqueue; ls qf* | cut -c3-)
do
echo "Processing message $i"
sendmail -qI${i}
done
