Ask and answer these questions about the cluster being upgraded:
- How many nodes are in the existing cluster?
- How much data is represented within each node of the cluster?
- Do we have enough equivalent hardware to reproduce every node?
- If not, can we subsist on a smaller server count for any duration?
- If not, is there sufficient storage capacity on existing cluster nodes to host two copies of the data?
- Do applications currently communicate through a proxy layer?
- If not, would it be difficult to introduce a proxy layer?