caipe-io Cluster Resource Health
Query AWS EKS clusters for node health, pod status, resource utilization, and alerts to produce a cluster health dashboard.
Instructions
Phase 1: Cluster Overview (AWS Agent)
- List EKS clusters and their status
- Check Kubernetes version - current vs. latest, end-of-support date
Phase 2: Node Health
- Inspect node conditions - Ready, MemoryPressure, DiskPressure, PIDPressure
- Resource utilization per node - CPU, Memory, Pod count
Phase 3: Pod Health
- Identify problematic pods - CrashLoopBackOff, ImagePullBackOff, OOMKilled, Pending
- Namespace-level summary - pods running, pending, failed per namespace
Phase 4: Resource Capacity Analysis
- Cluster-wide utilization - total CPU/Memory requested vs. allocatable
- Capacity risks - nodes at >80%, namespaces exceeding quotas
Output Format
```markdown
Cluster Resource Health Report
Cluster Summary
| Cluster | Version | Nodes | Status | Overall Health |
|---|---|---|---|---|
| prod-us-west-2 | 1.29 | 12/12 Ready | Active | HEALTHY |
Resource Utilization
| Resource | Requested | Allocatable | Utilization |
|---|---|---|---|
| CPU | 38 cores | 48 cores | 79% |
| Memory | 96 Gi | 128 Gi | 75% |
| ``` |
Examples
- "Check the health of our EKS clusters"
- "Are there any failing pods in production?"
- "Show me cluster resource utilization"
- "Which nodes are under memory pressure?"
Guidelines
- Check all clusters unless a specific cluster is requested
- Flag any node above 85% resource utilization as a capacity risk
- For CrashLoopBackOff pods, suggest checking logs as the immediate action
- EKS version end-of-support should be flagged at least 90 days before EOL
- Use kubectl read-only commands only (never modify cluster state during health checks)