Hello, We are in the process of isolating multiple applications currently running in one large cassandra cluster to individual smaller clusters. Each application runs in its own keyspace. In order to reduce/eliminate downtime for a migration, I was curious if anyone had attempted the following process to migrate data to a new cluster:
1) Add new cluster nodes as a new datacenter to existing cluster 2) Set RF for specific keyspace to non-zero for new cluster, use nodetool rebuild on new nodes to stream data 3) Change application node connections to point to new cluster 4) Set RF to 0 for original cluster (stop new writes from going to original cluster) 5) Break connection between nodes so new nodes become a standalone cluster??? <- Is this possible? what would be the high level steps? If this is an extremely bad or misinformed idea, I would like to know that as well! I am aware of other tools available including sstableloader, etc, but this seemed like a more elegant solution, leveraging cassandra's active-active features. Thanks, -Andrew NISC
