Originally from the User Slack
@Hans: Hey, I have a question about nodetool cleanup.
We have 3 nodes in each of DC1, DC2, and we need to replace a disk on every nodes in DC1,
my understanding is that we have to
• Decommission the node, replace the disk, recreate the scylla dir, and join it as a new node.
• Once the new node is added to the cluster, run nodetool cleanup on all existing nodes.
My question is:
Since cleanup just removes outdated data from SSTables, is it sufficient to run it once on each node after all rolling replacements are completed?
Or do we need to run cleanup on existing nodes after each individual node replacement?
Would appreciate any insights! Thanks 
@stewart: Replacing a disk and doing a nodetool cleanup are different operations. Which one are you trying to do?
@Samuel_Hameau: you may just stop scylla on your node ; copy the disk content to your new disks ; replace disks ; restart scylla with the new disks
@Hans: Thanks, everyone!
Stewart gave me the similiar advice as Samuel said.
Just drain → turn off scylladb server → replace disk → turn the scylla server back on.
This approach is much better than decommissioning and rejoining the server to the cluster.