0

I have two dedicated servers on OVH with ESXI hosting multiple VMs. VMs are multiple application servers and databases (one is active, and another is standby). Each server also has Apache mod_jk Load Balancer that balance between application servers on both dedicated servers over vRack.

Problem is that both Load Balancers have public IP, and I need only one active at a time.

I have a few approaches, but all of them have some obstacles.

Approach 1:

I tried to define multiple A records on IONOS, but according to some sources, there is a problem with caching, especially with Google Chrome.

Approach 2:

IP groups on both servers are Failover so I can switch IP to other LB in case of failure, but the switch has to be done manually (as far as I know).

Approach 3:

OVH has an IP Load Balancing option, but it still creates a Single point of failure. Also, documentation is not very helpful. Can I define rules to use backup LB only in case primary LB failed?

Is there some solution to these problems or any other approach?

Zivan Radosavljevic
  • 13
  • 4

1 Answers1

0

Approach 3 is the nicest one because it is integrated with the rest if the OVH Ecosystem. That offer is based on a custom implementation on HAProxy and is a highly available solution with multiple stage of loadbalancing (datacenter -> rack -> server -> HAProxy instance). You'll find almost all the basic loadbalancing from HAProxy (L7, TCP, SSL deciphering, etc. The product is quite easy to use.

Approach 2 may be an option if you're not bothered to do the switch manually. However you could try to automated it using tools like corosync, pacemaker, etc.

May I propose an fourth approach ? :) You could deploy a new VM with an HAProxy instance on it and set it up in VMWare Fault Tolerance mode ? That way, your VM is "live-duplicated" between two esxi ans a failover occurs automatically in case of an issue.

Maiko Bossuyt
  • 165
  • 6