Previously, Cassandra nodes were started concurrently (all at once). However, this is not a correct approach - the nodes should be started one by one, waiting for each node to fully bootstrap.
This commit changes the behavior to match those best practices. To determine whether a node has fully bootstrapped, checking the CQL port for availability is performed (in both Cassandra and Scylla, CQL port is only started after node bootstrap is fully finished).
Partially addresses #10 (Scylla logic is still wrong - I will address it in next PR dependent on this PR).
Previously, Cassandra nodes were started concurrently (all at once). However, this is not a correct approach - the nodes should be started one by one, waiting for each node to fully bootstrap.
This commit changes the behavior to match those best practices. To determine whether a node has fully bootstrapped, checking the CQL port for availability is performed (in both Cassandra and Scylla, CQL port is only started after node bootstrap is fully finished).
Partially addresses #10 (Scylla logic is still wrong - I will address it in next PR dependent on this PR).