Vicente Aguilar <[EMAIL PROTECTED]> writes:

> That's the point (no pun intended). We want to configure this backup
> system in order to avoid having a SPOF on the server without moving the
> SPOF somewhere else (the NFS server in your scenario).

Right, but we're using a NetApp which has oodles of redundancy built
into it already.  We hope to get another down the road so we can
cluster them.  That should eliminate the NFS SPOF.

> I still haven't taken a look at the source code, at this point we're
> only trying to figure out the best way to do this... but as far as I
> understand how the qmail-ldap clustering works (and I've never used it,
> just read some docs and this list), whenever a server in a cluster
> receives a POP/IMAP connection for a mailbox which is stored on another
> server of the cluster, it proxies the connection to the right place
> instead of doing local access to the mailbox. We would only have to
> modify the behaviour of all this so that all the implied processes do
> both things: local delivery/access _and_ proxying to the backup. It
> shouldn't be that hard to make a quick-and-dirty(tm) hack, providing
> that the code to do everything we want to is already there. Maybe
> Claudio could sched some light about this...

I was thinking the same thing: dupe the mail to the backup server with
QMQP, then dupe the mailbox changes by doing local changes plus
proxying the IMAP connections to the remote system.


> Of course we would also have to deal with errors on the connection with
> the backup, and provide some kind of mechanism (another queue?) to
> re-sync the backup whenever there's been a network glitch.

Yeah, that gets difficult. Plenty of challenges here.  :-)

Thanks.

Reply via email to