What problem does it solve? When Pods in a Huawei Cloud CCE cluster enter CrashLoopBackOff, ImagePullBackOff, OOMKilled, Pending, or Evicted states, engineers must manually correlate Events, logs, metrics, and node conditions to find the root cause. This Skill automates that read-only evidence collection and ranks the most likely causes with direct evidence. ## Core Features & Use Cases - Lifecycle-based cause ranking: Walks the Pod lifecycle funnel (admission, scheduling, volume mount, image pull, container start, probes, node pressure) and ranks causes such as ImagePullFailure, OOMKilled, SchedulingBlocked, StorageMountFailure, and SandboxOrCNIBlocked with cited evidence. - Structured evidence collection: Gathers Pod status, Events, current and previous container logs, metrics-server data, and node/PVC context through hcloud CCE and the kubectl-cce plugin, with strict cluster ID validation and credential redaction. - Scenario-specific guidance: Applies dedicated guides for ImagePullBackOff, CrashLoopBackOff, OOMKilled, Pending, storage mount failures, eviction, probe failures, CNI/sandbox failures, and quota/admission rejections, producing a Markdown or JSON report with ruled-out causes and handoff recommendations. - Use Case: A Deployment's Pods are stuck in ImagePullBackOff. The Skill resolves the cluster UUID, pulls Pod Events, identifies a missing repository tag, rules out scheduling and OOM causes, and recommends fixing the image reference—without running any mutating command. ## Quick Start Ask the AI to diagnose why Pod my-app-xxx in namespace default of CCE cluster <cluster-id> in region cn-north-4 is in CrashLoopBackOff and produce a root-cause report.
