Kamil Braun

And generally the upgrade procedure needs to be documented, with the endpoints that have to be POSTed and GETted and the waiting that needs to be done etc.

Created https://github.com/scylladb/scylladb/issues/17341 which needs to be done by dev team

Another thing we should document: in removenode and replace, we deprecate the option of passing IPs, you should be passing host IDs instead. (The option stays for now but will...

> One thing we must remember is documenting the operator's responsibility for shutting the node down in case when the topology coordinator fails to deliver the join rejection. This change...

We need to document that resharding in raft-topology mode requires quorum of nodes to be alive @vladzcloudius heads up -- you asked to communicate such things

You mean like a placeholder page to be filled later?

> We're already using host ID. https://opensource.docs.scylladb.com/master/operating-scylla/nodetool-commands/removenode.html and there is a warning about deprecating the IP use. It says: > However, this feature is deprecated with consistent topology updates enabled...

IIUC all required things were done in 6f374aa7d68e297b2f69df208a41eedf15507a15. The upgrade/recovery procedure has a separate issue: https://github.com/scylladb/scylladb/issues/17341 so we can close this one.

So apparently the fix did not address the original report -- which was about `test_simple_kill_node_while_decommissioning` -- instead it addressed all the other reports which were posted in comments -- about...

> Currently adding new service level throws Failed to apply group 0 change due to concurrent modification but the change is applied and it's visible on all nodes. I need...