Server Fault is a question and answer site for system and network administrators. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Are there strategies or techniques to create HA environment when IP failover isn't available? I'm reading a guide at Linode's library which describes exactly what I want to do, but I'm not sure my current VPS host supports IP failover.

share|improve this question
up vote 3 down vote accepted

IP failover using hearbeat/pacemaker requires the ability to move an IP address between nodes. If your VPS provider locks you into a single IP address, then that isn't the route you want to take. You'll need something where the IP address is stable.

Depending on what you're doing, something like may fill your needs. You place a server running a very simple service in front of a pair of server (or more), and clients talk to the haproxy server. Should a failover be needed, the haproxy server flips its back end communication to the still-working server. It does introduce a single point of failure, but it's a fairly simple service which may very well be more stable than the application you're hoping to make highly available.

HAProxy is just one of several ways of doing this, but it is a software-only load-balancer which is much more compatible than a hardware load-balalncer like an F5 device.

share|improve this answer
You can have multiple (stand-by) ha-proxy instances which replicate the global settings (and other global state like cookies etc.) and the stand-by is monitoring the first and uses VRRP. You talk to a shared IP then. See the haproxy architecture guide for that setup. – pfo Dec 13 '11 at 16:46

You could use Round-robin DNS.

The solutions that would be viable really depend on what the application is though.

share|improve this answer
Round robin DNS is good for load balancing, but not necessarily HA. – Matty Jun 12 '11 at 9:20
If nothing else he could move the RR record. Of course that would probably mean about 15minutes downtime (I've seen to many ISPs ignore TTLs lower than 15min and just overwrite it so I don't go lower than that) – Server Horror Jun 12 '11 at 10:14

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.