Check cluster health?
Make sure cluster is not in a degrade state?
Check if range replication status allows for it (i.e can survive a full failure)
On any node:
Adding a node first provides an opportunity for maintaining SLAs by keeping a sufficient number of vCPUs in the cluster available for workload processing.