Kamil Braun

@aleksbykov I see in https://github.com/scylladb/scylla-dtest/pull/3959/files there is: ```python3 ip4 = get_ip_from_node(node=node4) node2.watch_log_for(f"{ip4} gossip quarantine over") ``` I tried the same for this test: ```diff diff --git a/bootstrap_test.py b/bootstrap_test.py index c7c91f99..16ae8730...

`consistent_cluster_management` is still using "gossip-based replace", it only adds an additional step to the procedure, which is modifying group 0 configuration. Raft-based topology operations are only enabled under the experimental...

> With consistent-cluster-management the replacing node, when ignoring two other dead nodes, after cluster restarts. > Goes into a loop like this one: How many nodes are there in the...

> It would be clearer if group0 would have printed a clearer message to the log that it's waiting for a quorum of the nodes. Agreed, but nobody sent a...

> The rest of this series deals with loading the saved states also from storage_service::on_join for all token owners and also in the replace path, from storage_service::prepare_replacement_info. By "all token...

I put it in backlog, it shall be looked at eventually

@aleksbykov in raft-topology mode, when you call `removenode` on a node, it is permanently banned from the cluster. The same for nodes that you pass to `--ignore-dead-nodes`. Even if `removenode`...

@avikivity who should we assign this to? @scylladb/releng ?

@margdoc in your last post, you meant not `scyllaadm`, but whatever was passed to the `admin-group` option, right? --- @avikivity > I see scyllaadm is outside the scope of dist/debian...

Fixed by https://github.com/scylladb/scylladb/pull/17701