>> We use spread/wackamole for failover, and it works really quite
>> nicely, though not perfectly. I'd estimate that downtime is reduced by
>> an order of magnitude.
>>
>
> In my case I have a separate server for people to get their mail from.

You misunderstand me. I'm talking about SMTP services, which I regard as 
(a) for incoming email, and (b) for local mail submission. For local mail 
submission, high availability means providing high uptime for each IP 
address that services the submission, because mail clients won't fall back 
to a secondary IP address, even if several are available after DNS 
resolution.

> Incoming email goes to the front end spam filters, the spam is removed,
> and the good email is forwarded to other email servers where the users
> get their email.
>
> BTW, Ian, What do you use for a pop/imap server that runs across several
> computers?

We don't provide POP. As a campus service, we think its important to hold 
mail on our servers so that we can back it up, and so on.

We use Cyrus IMAP, with four mailstores on SAN storage devices. The SAN 
mirroring provides filesystem redundancy here. We're half way through 
deploying this, moving from UoW IMAP. The Cyrus IMAP server itself will 
eventually be clustered, but we're currently running a single IMAP server.

During the mailbox migration process, we used Perdition IMAP proxy servers 
so that we could migrate users without requiring them to change their mail 
client settings.



-- 
Ian Eiloart
IT Services, University of Sussex
x3148

-- 
## List details at http://www.exim.org/mailman/listinfo/exim-users 
## Exim details at http://www.exim.org/
## Please use the Wiki with this list - http://www.exim.org/eximwiki/

Reply via email to