kubernetes-operatorslisted

# Kubernetes Operators — CRD and reconcile review For pod-level debugging (CrashLoopBackOff, ImagePullBackOff, scheduling failures) see the `kubernetes-operations` skill. For Argo-managed deployments see `argocd-operations`. Most "operator bugs" are reconcile-loop bugs, not Kubernetes bugs: missing finalizers, blocking calls, no requeue on transient error, status drift, RBAC over-grant. The three Python scripts shipped here catch the deterministic subset before code reaches a cluster. ## When to invoke Run the analyzers first — they're stdlib Python, fast, and surface most routine issues: ```bash SKILL=plugins/kubernetes-skills/skills/kubernetes-operators python3 "$SKILL/scripts/crd_validator.py" --crd config/crd/ python3 "$SKILL/scripts/reconcile_lint.py" --controller controllers/ python3 "$SKILL/scripts/operator_capability_audit.py" --operator-dir . ``` All three accept `--format json`. Triage by severity: FAIL blocks merge, WARN files an issue. ## Pre-flight: is an operator the right shape? Operators are for *stateful, lifecycle-managed* workloads. Reach for one when: - The thing being managed has an external API (RDS, Kafka topics, GitHub repos). - Day-2 operations are non-trivial (backup, restore, version upgrade, failover). - A Helm chart + bash isn't enough — you need a controller that *observes and re-acts*. Don't reach for an operator when: | Want | Better tool | |---|---| | Run a workload | `Deployment` / `StatefulSet` / `Job` | | Pa