1

Is it possible to set up Remus to act as a high availability server across 3 locations. Has anyone set up something like that. That is A would be main site, which would have an HA set up at B, and C should take over both A and B go down.

I have seen successful implementation of Remus for site to site. But have no idea how site-to-site-to-site will work.

1)How will you handle split-brain conditions in such cases? 2)Or is there an alternative in FOSS world to Remus?

ramdaz
  • 685
  • 1
  • 10
  • 26
  • Depends on what you are clustering there. Remus is awesome, but IMHO still beta. A simple failover-cluster is much easier - but will loose connections. More complex would be an application-server-cluster based on Tomcat with Tomcat-means... – Nils Sep 21 '13 at 21:07
  • I have been running VOIP servers on Remus for past 3 years without any hiccupos, but that's site-tosite. I have no clue how to do site-to-site – ramdaz Sep 22 '13 at 07:40
  • VOIP is `Asterisk`? Apart from that - how about firing up C taking over the role of A after A is down and B has taken over? – Nils Sep 22 '13 at 21:07
  • No. VOIP is not Asterisk, but from a German vendor called SNOM. It would work with Asterisk too. Nils, a good idea, but how do you fire up C, since Remus works machine to machine transparent HA. – ramdaz Sep 24 '13 at 20:44

1 Answers1

0

The idea is to have the third node as warm standby.

It should have a relatively synchronous disk-replication in place (perhaps with stacked drbd-devices?).

Apart from that it will take over the role of A or B including IPs and Remus-Setup if A or B fails.

Nils
  • 7,695
  • 3
  • 34
  • 73