We have Cassandra 4 node cluster at AWS and 4 Nodes at Azure with RF 2 on each DC. We want to use local_quorum consistency level for for strong consistancy.
Configuring Apache Cassandra for Disaster Recovery
Is it mandatory to change RF to 3 at Primary DC as mentioned in the above thread?
The choice of the RF is about consistency and also about resilience.
For the QUORUM consistency level you need that at least more than half node that hold a data answer to a query.
With RF = 2, request with QUORUM consistency level need an answer of the two nodes that holds the data corresponding to the request. So, if one of your nodes goes done, you won't be able to get the QUORUM consistency level for an important range of your data.
With RF = 3, if a node goes down, the two others will guaranty the availability of the QUORUM consistency level.
