Re: Building redundancy
Oliver Leitner <shadow333 <at> gmail.com>
2005-06-01 00:15:51 GMT
On 6/1/05, Ashley M. Kirchner <ashley <at> pcraft.com> wrote:
>
> First, the question: How to build redundancy on a "users" server?
>
> What's behind it: When our main users server goes down, no one can
> log in to check their e-mail. So the problem as it was presented this
> morning was to research some way to make it so that if that server goes
> down, users still have a way of getting to their e-mail or login to
> their accounts for that matter. Now, the setup we have is a bit
> different from most people I'm sure. Incoming e-mail flows as follows:
>
> Internet -> MX server -> spool server (NIS+ slave)
>
> Our "users" server - also our NIS+ master - (which is a totally
> different machine) does an NFS mount of the spool server on /var/mail/
> and voila, e-mail. When our users use imap/pop to check/get/send
> e-mail, they log in to that "users" server to do so. No one
> communicates directly with the spool server. So when the "users" server
> goes down, no one can get to their e-mail. The question: what kind of
> redundancy can I build so that if that server were to go down, that
> users can still log in (to -something-) and still access everything they
> need. Presumably this would be a separate machine and that the fall
> back would be transparent to them. I just don't know how or what.
>
besides the obvious (keeping the server running) id consider you think
about a mirroring type of system, i am not sure if anyone has yet
tried something like that for mailservers, at least for webservers its
pretty common to have 2 or 3 or more servers, who are updating their
content with each other at special given times.
(Continue reading)