I tried posting something similar in networking but this is more of a general question now: How does failover between two linodes occur? I tried setting up keepalived and then wackamole/spread, neither worked. I found this was because both use multicast addressing for messaging between nodes, and Linode only supports unicast addressing.

Is there a utility that can use unicast for failover between multiple machines, or is there a way to set up keepalived or wackamole/spread to use unicast messaging? Would heartbeat work for this?

Has anybody set up failover between two linodes (on the same internal network)? How did you do it?

I really appreciate your help!

I got it to work by using a GRE tunnel to send the multicast messages between nodes. I'm using ip_vs, keepalived, and 2 nodes.

I had no idea GRE tunnels existed until now. I'll do some more research. Thanks for your help!

PPTP VPNs use GRE tunnels.

I'm also thinking on buying a second node to try to have some sort of HA...
can you tell me if your second node is located on the same datacenter of the first node please?

Relevant:

There have been a couple good HA library articles posted recently.