- Shipped
- August 4, 2026 at 10:41 PM UTC
- Author
- Kamo
- Commit
- 14d9005
This instance is kamo's: it backs the `kamouniverse` keyspace (read/written by **************** through the scylladb-direct ExternalName) and the `audit` keyspace. The `ratestack-` cluster name was a leftover from before the estates were consolidated and actively misleading — during the ratestack teardown it read as ratestack-owned and nearly got reclaimed along with 8 CPU / 32Gi. Renamed live using the required order (UPDATE system.local + nodetool flush system on the running node, then change --cluster-name and restart); Scylla refuses to boot if the two disagree. Verified after restart: ring UN, all keyspaces present, row counts unchanged, universe-persistence reconnected and writing. Also brought this file back in line with what is actually deployed: - NetworkPolicy: drop the now-torn-down `ratestack` namespace from the allow-list. - Probes: the repo still carried the old aggressive values (exec readiness, 60s/6 liveness) while the live object had been relaxed to tcpSocket + 900s/80. Adopted the live values so an apply no longer regresses them — the stale ones also made `kubectl apply` fail outright with a two-handler probe merge. - README/comments: said k3m1 and described a StorageClass/PV layout that was never used; it is k1m1 with hostPath mounts. Documented the rename procedure and why grepping for consumers of this Scylla finds nothing.