MailPlus HA + disaster fault tolerance

[verdog123]

Hi everybody. Nice to meet all of yours, this is my first post. Hope all can colabborate with my situation.

Im looking for a production implementation of mailplus server. I have been looking for information and best practices about that. Obviously the first thing that come to my mind when thinking in a fault tolerance escenario its to set up a HA cluster. Thats very clearly to me.

The problem comes when i think in a disaster recovery, such as the site A (where reside the cluster) get a faultly of internet or energy supply for 3 or more hours.

I'been thinking in a Site B, connected by LDAP and where all the mailboxes are already syncronized and updated when Site A shutdown and become inaccesible for clientes. I think to Setting Site B (in mx dns of my domain provider) with a high number prority than Site A will resolve the mail reception, but the sync of mailboxes its my big problem.

Did anyone the same setup? What do you think about connect 2 synology via l2tp and simulate a "local area" connection to perform a HA cluster through VPN?

Thanks for your answers.

Verdog123

 

[Rusty]

Site to site replication can be done using Snapshot Replication package. So you could have a "cold" mail setup ready on your site B location with data being replicated there.

If you need to initiate a failover due to site A being down, you can do it from site B even (in case site A is out of power) and just re-protect data on site B and boot up your mail system. Ofc some DNS changes will need to be made, but data replication will not be an issue.

 

[verdog123]

Thank you Rusty for your reply.

I will test that setup. In DNS i will put Site A static IP with 0 priority and Site B with static ip and priority set to 1. That is enought to properly mail delivery in site A unless its down?
I want (if something happens to cluster site A) to automatic failover mails to site B. So when site A come back to normal activity mails will change again to delivery in site A. How Snapshot replication run in this situation?
When A fails, maybe i need to let the cluster with 5 priority, re-configure the snapshot replication, this time from B to A and then do the failover to maintain all of those mails who was delivered in site B when the A was off?
Hope you can understand me.
This is a setup with 70 clients and its important that mails can be exchanged everyday, so i cant lose any mail and cant let any windows-time with the mailsystem off

Thanks to all!
Regards,
Verdog123

 

[Rusty]

How Snapshot replication run in this situation?

It will not autofail, you will have to do the manual handover and reprotect. Its a backup solution, not so much a HA solution. But it does work great with "live data" so that's why its good for such situations.

When A fails, maybe i need to let the cluster with 5 priority, re-configure the snapshot replication, this time from B to A and then do the failover to maintain all of those mails who was delivered in site B when the A was off?

I get you yes, but again, failing back you will have to have side B configured as main (due to Side A being down), and once you recover site A you can then initiate again manual failover back to site A as your main storage and activate mail service on that side.

This is a setup with 70 clients and its important that mails can be exchanged everyday, so i cant lose any mail and cant let any windows-time with the mailsystem off

maximum replication time gap for Snapshot is 5min. So every 5 min you can have snaps executed between sites. How fast will you react when your site A goes down and switch it to site B will be up to you.

The only auto-switch/failover scenario is when using HA cluster (on premise), but in a case of location B (off-site) you will not be able to use SR tool for auto-failover.

One way you can try is to use site to site VPN and configure a streched HA setup in the same network subnet between 2 locations. Maybe in that situation you can have auto-fail if HA will not detect this as 2 remote locations. I haven't had a way to test this.