The PriorityClasses exist in the repository but no pod uses them, so every platform pod still evicts at priority 0. Attach platform-critical to Keycloak (CR spec.scheduling) and Gitea (chart priorityClassName), and platform-low to Loki (singleBinary) and Tempo. The Keycloak CR also gets spec.image at the running 26.7.0 digest with startOptimized: false: with spec.image the operator otherwise assumes a pre-built optimized image, and this is the stock one. Grafana and pgAdmin have no top-level priorityClassName in their charts and are left for later; platform-postgres is W1's file and tech-log belongs to its deploy/ directory. Risk: applying restarts Keycloak, Gitea, Loki and Tempo once, so it belongs in the k3s restart window, after the PriorityClasses exist (a pod that names a missing class is rejected). The accepted-drift list holds these objects until 2026-11-30. Verification (desktop): render-phase1/2 and render-observability-core pass with the four priorityClassName values; kubectl diff (server dry-run) against live shows only these fields.
Loki
This root renders grafana-community/loki chart 18.7.1 as one Loki
3.7.4 monolithic StatefulSet in observability.
- The image is pinned to the reviewed amd64 digest.
- Logs use the
lokiAIStor bucket throughloki-object-storage; no root credential is referenced. - Retention is 30 days. The WAL and working data use the dedicated 5Gi Retain Local PV.
- Gateway, canary, caches, embedded MinIO, tests, and rule sidecars are disabled.
- The only client-facing service is ClusterIP port 3100.
Do not delete the PVC or bucket during rollback. Scale the StatefulSet down, preserve its Retain volume, and investigate S3/WAL state before reapplying.