If I set the replication configuration to “three_data_hall” and also set the “locality_zoneid” on each machine to the same value as “locality_data_hall”, will the replication constraints be able to be satisfied?
There are enough machines in each data hall to satisfy the constraints before setting “locality_zoneid”, but to make some operational scripts easier for using both triple and three_data_hall in different clusters using the same tooling, we would just like to always set zone ID to avoid templating based on the replication configuration.
No, I don’t think that will work. The replication for the TLogs in three data hall is 2hall x 2zone, so we keep log replicas on two different zones in each data hall, and in two different data halls. This means each data hall needs at least 2 zones in it to have a valid TLog team. Ideally you will want 3 zones in each data hall.
For context, the goal in this configuration is that the database should be able to survive the failure of 1 entire data hall and 1 zone in another data hall, without losing durability or (with sufficient additional zones) availability, which is what drives the zone requirements.