[ILUG] Secondary mail servers

[Enda]

David Wilson Wrote:
> Rick Moen wrote:
> >  I also realised that there's no disadvantage to eschewing backup
> >MX as long as you're confident you can bring up _some_ replacement mail
> >service for your domain, somewhere in the world, within four days of
> >your primary site's failure.  (You would need to re-point the DNS, of
> >course.)
> >
> >Four days is the defacto standard for MTA delivery time-out periods.
> >
> >
> That is indeed the case for many default mail server installations,
> however you'll find that in the backup MX market, the average timeout is
> much higher for obvious reasons.
>
> You can't run a single mail server for a mail domain and expect anything
> good to come of it. It is also highly irresponsible if failure of the
> domain affects more people than just yourself.
>
> There are many factors that you must draft in when configuring a mail
> domain beyond how fast you can configure a replacement mail server, for
> instance, say you've been hit by a bus - is there someone else on-site
> who can perform the task? Failing that, how happy would your employer be
> to know that the weakest link in their mail system was you?

Lets face it, a backup MX doesn't solve any of those problems anyway, hit by
a bus or not, someone still has to get a replacement server up and running
for the services to return. All you're doing is caching mail for forwarding
instead of letting the originating MTA cache. At least in the later, after 4
days the original poster gets notified, and can contact your employer who
will otherwise be clueless to the problem. Delivery to a backup mx in the
absence of the primary being restored yields a false delivery without the
sender being notified.

All you've managed to do is prove that backup mx's are useless for
resilience and that the future is clustered primary services with load
balancing or round robin routing.

 -Enda.