All the mail for our Unix service is delivered to one host, a SparcCenter 1000
with 6 processors and 512Mb memory, currently running sendmail v8 (until I put
exim on it). Due to the volume of mail, whenever this host comes back up after
downtime, exim on our campus mail relays rapidly delivers all its pending mail.
This is OK except that the host is then swamped with hundreds of sendmail
processes all doing asynchronous delivery (option Odb is set): the load average
has exceeded 170 at such times!
My boss thinks exim ought to have some sort of throttle control to prevent
this. I disagree. I pointed out to him what I thought was exim's approach, that
if the remote host was accepting the mail, it wasn't exim's problem, it was the
remote MTA which was poorly configured. Nevertheless, I told him I'd ask if
it's possible to somehow throttle exim back from being quite so fast, for
certain hosts :-) I know about the batch_max option on the smtp transport,
which I could use in a special case. He asked if it's possible to manually
freeze all mail for a specific host, i.e., to tell exim to behave as if a host
is down (even if it isn't), in cases where that host has problems affecting
mail delivery.
It seems fruitless to me, given that I plan to upgrade the host in question
from sendmail to exim (the difference in invocation of procmail between
sendmail and exim is the only problem), but there's management for you. At
least I've asked.
cheers,
David
--
David Osborne e-mail: Postmaster@???
Postmaster tel: +44 (0)115 951 3397
Cripps Computing Centre
University of Nottingham, UK
