High Availability - Email, specifically.

[InTheWoods]

I've searched the documentation as well as the web, with forum posts from 2009, 2013, and other times of current irrelevance being displayed. Documentation does not seem to mention "high availability", but does mention clustering.

I'm curious if anyone here has setup a High Availability solution, specifically relating to email. We'll have two separate physical servers with internal private networking. In the event one goes down (maintenance, upgrade, hardware failure, etc) we'd like to keep things humming along as usual by automatically failing over to the machine that is working.

Can anyone share some experiences with a similar setup? I'm reading a bit here: https://docs.directadmin.com/directadmin/general-usage/multi-server-setup.html but not sure it is exactly what I'm looking for.

Basically, just have some higher paying business clients who want to always have access to email and despite the added cost in an HA setup, it seems worth it to keep them happy since they're paying for it.

 

[DanielP]

second one no need to be Directadmin just dovecot + Exim is enough and maybe webmail

Replication with dsync — Dovecot documentation

Two way Dovecot Server Replication / Mirroring - Techy Help

Step-by-step guide for two way dovecot server replication / mirroring without compromising server performance. Apply this and achieve replication in no time

shoreup.co.uk

P.S In my experience I saw maybe over 10 times more network events than hardware events so replicating in same datacenter is almost meaningless choose a different dc or get some small vps from cloud provider that is not in same area ...

 

[InTheWoods]

second one no need to be Directadmin just dovecot + Exim is enough and maybe webmail

Replication with dsync — Dovecot documentation

Two way Dovecot Server Replication / Mirroring - Techy Help

Step-by-step guide for two way dovecot server replication / mirroring without compromising server performance. Apply this and achieve replication in no time

shoreup.co.uk

P.S In my experience I saw maybe over 10 times more network events than hardware events so replicating in same datacenter is almost meaningless choose a different dc or get some small vps from cloud provider that is not in same area ...

Thanks, will check that out.

My experience has been the opposite, but we have our IPs announced in a few different locations so I guess a regional failover would be better regardless since the option is available and would be a stronger selling point. I've just seen less total network outages than I have seen "Oh crap, a motherboard just died with no warning signs" on some of these Epyc builds, ha. But in any case, having two separate geographic locations would address both concerns and shouldn't cost more or be more difficult to manage/setup than having it all under one roof.