- Ya
- 4 Agosti 2026, 22:49 UTC
- Mwandishi
- Kamo
- Ahadi ya
- 9ca1f2c
scylladb/ held complete manifests but had no deploy step — the comment said it was "deployed MANUALLY after the 16TB drive swap" because the data dir /mnt/backup/scylla had to be physically on k1m1 first. That condition is met: the drive is on k1m1 and Scylla is running from it, so the manual exception no longer applies. Adds a scylladb/-gated step applying service, networkpolicy, poddisruptionbudget and statefulset, matching the change-scoped pattern used by the other components. Note this commit touches .forgejo/, which forces FULL=true — a full-cluster reconcile. That was verified safe first by diffing all 108 apply targets against live: 105 were identical, the 3 that differed are bootstrap Jobs the workflow deliberately deletes and recreates, every kubectl delete target is absent (no-op), and the *** secret rebuilt from repo files hashes identical to the live secret. The two manifests that would have regressed live state are fixed in the preceding commit.