Lukasz Sojka

A reminder for this bug - as we still set this value for GCE and Azure preinstalled images. If this setting is bogus, we should stop doing that. cc @slivne

Got another reproduction: ## Installation details Kernel Version: 4.19.0-21-cloud-amd64 Scylla version (or git commit hash): `5.0.3-20220907.b9a61c8e9` with build-id `7be266d2954825cdf843c744de04a0443a8f156c` Cluster size: 4 nodes (n1-highmem-8) Scylla Nodes used in this run:...

Issues from both comments are not related to issue opened by @k0machi - this is first run of this nemesis and fails on first c-s upon creating keyspace across multi-dc....

The same happened during `destroy_data_then_repair` disruption. More test details: ## Installation details Kernel Version: 5.13.0-1025-aws Scylla version (or git commit hash): `5.0~rc6-20220523.338edcc02` with build-id `60217f35371db2b1283e0c5bc67a7f5604768d41` Cluster size: 6 nodes (i3.large)...

@roy There were no changes to that test recently. I also tried to reproduce by running this scenario in a loop for few hours on my laptop but didn't reproduce.

I was trying to reproduce it but didn't make it. I'm not aware of reproduction in any other test run. Maybe this was one-time issue related to something else (e.g....

I think we can close until reproduction.

> @soyacz do you have a stuck node ? are you able to reproduce ? > > If we have a stuck node in this state we need complete logs...

I got reproduction: after upgrading 2 nodes and trying rollback one node to 5.0.3 from `2022.1.3-0.20220920.941a34e52 with build-id d9918d99b8935828ce27f46d2a773490607b14fb` drain freezed. From logs I can see last line is: ```...

Machines are getting terminated after the test. SCT should collect coredump on timeout in this case: `node.run_nodetool("drain", timeout=3600, coredump_on_timeout=True)` I'll try to fix it.