At least you need to prepare

1. 2 Frontends for the cluster (Of course!) each one has
   1. 1 hostname.
   2. 1 Public IP connecting to outside world.
   3. 1 Private IP (each with different IP) connecting to private network
2. 1 Public IP as a central public address for the cluster
3. 1 Cluster hostname as a central public hostname for the cluster, which will be associated with Cluster public IP.
4. Also another 1 private IP as a central private address for compute nodes.
5. NAS or anything that will act as Cluster file system. We will depending on storage tier to taking care of /home and /share space synchronization. We will not deal with how to synchronize file between Frontend here, although it might be possible with DRBD.

For software requirement

1. DRBD
2. Heartbeat
3. rsync, ssh
4. Some cluster file system, this can be either plain NFS or NAS server, or complex Cluster FS like Lustre or GFS. Please note that, for my setup we use Lustre here.

Assume a frontend to act as primary frontend. I will name this one FE1 from now on. The secondary frontend will be named FE2. Then plan your configuration and put it in paper. Pinned it to the wall. For my configuration, it is

fe1

• private IP - 172.16.0.2
• public IP - 203.123.123.2

fe2

• private IP - 172.16.0.3
• public IP - 203.123.123.3

Central IP address (shared among 2 FEs

• private IP - 172.16.0.1
• public IP - 203.123.123.1
• name - cluster

NOTE: I have to use IP range 172.16.0.0/255.255.0.0 instead of default 10.0.0.0/8 because the 10.0 was taken by internal enterprise network. I think it should be ok to use 10.0.0.0/8 instead.