Clustering — what needs to be

The last picture on the missing notes on the role of each node. Not really it will be clear who does what (MySQL, NAS, Apache/Nginx, Mail)

You can still ping the nodes in the plate.
And pull in addition to information about packages, info about hardware (for example, via lshw with the-xml) will be the inventory of iron.
\r
Info about packages and hardware is convenient to assemble in the background cron-om, and for the server to parse and put into database.
It is still possible for clusters to support wakeuponlan — to make changes in the init script to the client (to allow the "awakening" package) and to Wake from the server or "live" nodes in the grid.
\r
The same of the old hardware can then be used as variables for configuration: for example, the mac address setevoj: for example, you can make an array eth-addresses={"eth0": "00:11:22:33:44:55", "eth1": "11:22:33:44:55"}
\r
This is useful if the config write with inserts bash or another language.