Originally from the User Slack
@Akki: My current Scylla Cluster is 5.2 running on CentOS 7. Scylla 5.4 does not support C7, what steps I need to take to smoothly transition from C7 to Rocky?
@Felipe_Cardeneti_Mendes: Standard practice is to add new nodes and then decommission the old ones. Of course other ways exist.
@Akki: Adding new nodes is not an option. I was thinking for removing one node at a time rebuilding it with new RockyOS and Scylla5.4 and adding it back to the cluster. But my question is, as the rolling upgrade continues the cluster will have hosts with different OS. Will it be an issue?
@Felipe_Cardeneti_Mendes: You can’t bootstrap a node in a diff version as it introduces newer features. You can take it down and upgrade it/move around data then start
This is unsafe of course as if another node fails in between you lost quorum
@Akki: What would be the best way this can be handled Centos7 to RockyOS? (rolling preferred)
@Felipe_Cardeneti_Mendes: You should be ok with just postponing the upgrade for later. Upgrade first (remove, upgrade, add) to same version, then just upgrade ScyllaDB once complete
@Akki: kool, so do we update OS on seeds first?
@Felipe_Cardeneti_Mendes: Doesn’t matter much
@Akki: Sounds good, will keep posted