Doug wrote:
> On 5/24/05, Michael Sprague <mfs@???> wrote:
>
>>luqman kondeth wrote:
>>
>>>newbie question
>>>HI,
>>>Can anyone give me instructions on how to move exim server from one
>>>hardware to another. Basically we are dumping our old machines , so we
>>>need to move the exim from an old server to a new one.
>>>Thanks in advance
>>>luq
>>
>>Well, assuming you're running the same version of exim, it would be a
>>matter of copying over the configuration file and any support files to
>>the new server. Hopefully you'll be able to test on the new hardware
>>before dumping the old hardware. :)
>>
>>If you going to a newer version of exim, you will probably have to tweek
>>the config a bit to get it working correctly. If you're talking about a
>>major upgrade (exim v3 to exim v4), you should check out the website for
>>advice on how to convert.
>>
>>thanks,
>>mikeS
>>
>
>
> What if there are large mail spools involved? How would you go about
> migrating data so that there is as little downtime as possible (can it
> be done so that there is no downtime?)?
>
> The new box will have the same hostname and same IP (eventually). The
> new machine will be hidden behind a firewall and NATed if this helps
> any.
>
> I'm currently going through a number of possible ways out of this, but
> just wondered if others have worked through similar probs?
Just to be clear, by mail spools, are you referring to /var/spool/exim
or to actual mailboxes? If the latter, then the answer would depend on
what you're using as a mailstore.
If the former, then it would depend on how you can do the migration.
You mention the new server will eventually have the same hostname and
IP. After you put the new server in production, if the old server can
be assigned a different hostname/IP, then you should be able to flush
the queue until it is empty.
I believe you could copy the contents of /var/spool/exim to the new
server as well and it would pick up where it left off once exim is
started; but I'm not 100% sure on that. Plus if the queue us huge, this
may take some time.
In short, if we're talking about /var/spool/exim and not mailboxes, and
the old server can still be 'on the Net' after the new server is in
production, then you can do things with minimal downtime. A procedure
like this comes to mind (just a suggestion):
1) Build out new box and copy exim configuration over. Test like hell.
When you're satisfied the new box will work in a production environment...
2) Change the IP of the old server.
3) Give the new server the old server's former IP.
4) Start exim on the new server and watch email coming via the logs, etc.
5) Flush the exim queue on the old server until it's empty.
6) Recycle old server into something useful. :)
Downtime should only be a couple of minutes to change the IP of the
servers. Plus, if things don't work as expected, you an backout to the
old server.
Note that I'm assuming your exim server is basically just a relay and
not doing final delivery. If it's doing final delivery, you'll need a
take a good look at the software you using to store the mailboxes too
(which is generally not an Exim thing).
I hope this helps,
mikeS
--
Michael F. Sprague | mfs@???
Partner | System and Network Engineering (SaNE), Inc
use STD::disclaimer;
