Triage Kubernetes Pod Scheduling, Startup, and Readiness
Use Kubernetes events, owner state and logs to distinguish failed admission, Pending, crash loops, failed probes and service routing.
Confirm cluster, namespace, workload owner, revision and the authorization to inspect them. Start read-only with workload and controller state, then identify the exact Pod or note that none was created. An updated Deployment can still have older Pods serving traffic. Identify which revision and container exhibit the symptom. Stateful workloads additionally need identity, persistent volume, and ordered-update context. Do not delete or restart an instance simply because it appears unhealthy. ## Key Points - [Kubernetes Debug Pods](https://kubernetes.io/docs/tasks/debug/debug-application/debug-pods/) - [Kubernetes Pod Lifecycle](https://kubernetes.io/docs/concepts/workloads/pods/pod-lifecycle/) - [Kubernetes Resource Quotas](https://kubernetes.io/docs/concepts/policy/resource-quotas/) - [Kubernetes probes](https://kubernetes.io/docs/tasks/configure-pod-container/configure-liveness-readiness-startup-probes/) - [Kubernetes taints and tolerations](https://kubernetes.io/docs/concepts/scheduling-eviction/taint-and-toleration/)
skilldb get containerization-skills/kubernetes-pod-triageFull skill: 31 linesTriage Kubernetes Pod Scheduling, Startup, and Readiness
Establish context and controller
Confirm cluster, namespace, workload owner, revision and the authorization to inspect them. Start read-only with workload and controller state, then identify the exact Pod or note that none was created. An updated Deployment can still have older Pods serving traffic. Identify which revision and container exhibit the symptom. Stateful workloads additionally need identity, persistent volume, and ordered-update context. Do not delete or restart an instance simply because it appears unhealthy.
Separate admission and scheduling. If a replica never appears, inspect controller events and admission errors before looking for a Pod. Compare resource requests with namespace quota and limits, and confirm RBAC separately when an inspection or controller operation is denied. An existing Pending Pod can have scheduler evidence about node resources, affinity, selectors, taints, tolerations or unresolved storage. Inspect the actual messages and candidate constraints. A toleration permits scheduling under a matching taint but does not select a node.
Follow container and service state
Use the Pod description and container status to differentiate image pull, startup failure, repeated termination and readiness failure. Identify the specific app or init container and inspect previous container logs where available. Compare exit reason, restart count and timestamps with events. Sanitize logs before sharing. If the container is running, distinguish a startup probe, readiness probe and liveness probe. Readiness failure changes service eligibility, while liveness failures can restart a container. Use a probe that measures the intended condition rather than hiding the symptom through a longer threshold.
When the workload is Ready yet traffic fails, inspect the Service selector and endpoints for the intended revision. Check the configured target port and use the existing service mesh guidance when a sidecar or policy owns the route. If the app was installed through Helm, compare the chart release and values to rendered resources in the same namespace. Keep operator-reported drift separate from observed rollout configuration.
Produce a scoped diagnosis
Record the first failing stage, time, resource identity and supporting event or log. Propose one change to test in an authorized environment and name the observable recovery condition. Verify that the controller reaches desired replicas, the expected Pods become ready and a representative request succeeds. Include any remaining uncertainty such as missing retained events, unavailable logs or permissions. Record why a proposed restart would or would not address the evidenced cause.
References
Install this skill directly: skilldb add containerization-skills
Related Skills
Container Registries
Container registry setup, authentication, and image management for ECR, GCR, GHCR, and Docker Hub
Container Security
Container image scanning, runtime hardening, and security best practices for production workloads
Docker Compose
Docker Compose configuration for multi-service development, testing, and local orchestration
Docker Networking
Docker networking modes, custom networks, DNS resolution, and multi-host connectivity patterns
Dockerfile Optimization
Multi-stage builds, layer caching, and image size optimization for production Docker images
Helm Charts
Helm chart creation, templating, dependency management, and release lifecycle for Kubernetes