Shuttle Success Criteria
- Deploys to a DOKS cluster and registers with Starbase
- Creates the full per-customer resource set with correct labels from desired state
- Fluent Bit and vmagent successfully attribute logs/metrics via Shuttle's labels
- Reports pod snapshots every 60s for correct per-minute billing
- Reports cluster capacity every 60s
- Handles Starbase unavailability without crashing
- Garbage-collects resources removed from desired state
- Applies cluster_config changes (e.g., LB size unit)
- Passes
go build ./... and go vet ./...
- Memory under 200Mi for 500+ customer pods
Cross-references
The per-customer resource set in criterion 2 → §20.2
· the labels behind criterion 3 → §24.1 · the snapshot/capacity
cadence in criteria 4–5 → §25 · Starbase-unavailability handling
in criterion 6 → §27 · the memory ceiling in criterion 10 →
§28.