Once you initialize the root directory, you are ready to start the database using the voltdb start command. You issue this command, specifying the location of the root directory, the number of servers required, and one or more server addresses to use as "host" to manage the initial formation of the cluster. You issue the same command on every node in the cluster. For example:

$ voltdb start --dir=~/database     \   
               --count=5            \   
               --host=svr1,svr2         

On the command line, you specify four arguments:

You must specify the same number of servers and hosts (listed in exactly the same order) on all nodes of the cluster. You can, optionally, specify all nodes of the cluster in the --host argument. In which case, you can leave off the --count argument and VoltDB assumes the number of hosts is the total number of servers.

When you start the database, all nodes select one of the servers from the host list as the "host". The host then:

At this point, the cluster is fully initialized and the "host" ends its special role and becomes a peer to all the other nodes. If the database was run before and command logs or automated snapshots exist, the cluster now recovers the data from the previous session. All nodes in the cluster then write an informational message to the console verifying that the database is ready:

Server completed initialization.