Originally from the User Slack
@Hyunwoo_Kim: --------
one more question 
When there is 6-node cluster, and I need to replace the disks on 3 of the nodes. There seem to be two possible approaches.
- Decommistion and rejoin : Decommision the nodes, replace the disks, and then rejoin them to the cluster
- Drain and copy data : Drain the node, replace the disks, and copy the data back to the new disks.
Which approach would be more suitable in terms of data consistency, and overall?
@avi: Decommission + bootstrap is safer, since you have 3 replicas at all points in the process.
Drain + copy will be faster, but puts you at some risk in case another node fails during the process.
@Hyunwoo_Kim: Sounds interesting… Thank you for answering. I’m gonna find more info about Drain + Copy 