Re: [Exim] multiple exim, same spool

On Sat, 2004-07-10 at 01:29 -0600, Dan Egli wrote:
> explain lmtp?

it's basically SMTP for local use, the difference is that it presumes
the sending host is responsible for keeping the messages in a queue
until the recipient host is able to accept it. the key to making this
work is that you can defer messages per recipient after DATA.

the most common use of LMTP is probably to deliver to Cyrus.

> And what if it's 2+ machines sending mail? I'm not necessarily trying to
> justify sharing the spool, just coming up with situations where I would
> think of doing that, and trying to find alternates.

don't say spool, that's really not what we're discussing, the issue is
sharing the mail queue. sharing the spool is fine, we've had a spool on
NetApp with four delivery servers (running Solaris) for years with very
few problems.

sharing the queue is a bit different. the upside is that if one of your
sender hosts go down, some other sender host can pick it up and handle
it. if the queues were separate and local to each host, the messages
would be stuck until the host was back up. however, messages shouldn't
spend time on the queue, and if they are queued, not many will notice an
additional 30-45 minutes of downtime. yes, each sender host is a single
point of failure, but only for the messages in the queue. if you have a
shared queue, on the other hand, that shared storage becomes a single
point of failure for _all_ outbound deliveries.

--
Kjetil T.