Services without which nothing works are clear candidates for doppelgängers on my network. If the master fails, or if I just need to shut down the server for maintenance, I want the service automatically to start on the second server and to be available on the same IP address as before, if possible.

To do this, the IP address needs to migrate quickly and without much overhead to the backup machine. Addresses like this are known as floating IPs. The migration helpers here are Pacemaker [1] and Corosync [2]. These tools can't do much more than facilitate the move, however, so the solution seems a little over the top for a simple failover scenario.

A more streamlined solution called Keepalived [3] is part of the default toolset in most distributions. I just installed Keepalived and jumped in feet first. What I found were two servers with IP addresses 10.0.0.1 and 10.0.0.2. My floating IP of choice is 10.0.0.100.

Minor Difference

The /etc/keepalived/keepalived.conf configuration file looks like Listing 1. It differs on the two machines only in one aspect: the priority (line 16). I need to configure a smaller number on one of the machines than on the other to define the master. This server gets the floating IP by default, whereas the other only gets the IP if the master is down. The VRRP (Virtual Router Redundancy Protocol) is used to swap states.

To be able to bind services to an IP address that does not (yet) exist on the system, I need to make some changes to /etc/sysctl.conf:

sudo echo "net.ipv4.ip_nonlocal_bind = 1" >> /etc/sysctl.conf
sudo sysctl -p

Now I can launch Keepalived by typing:

sudo service keepalived start

The floating IP appears on the server with the higher priority value. If I shut down the master, I can see from the Syslog on the secondary machine that it quickly assumes the master role (Figure 1).

Figure 1: A glance at the Syslog on the slave machine shows that it has become the master after a failure of the production machine.