AGENTIC MEMORY

Skills

Useful instructions for your AI. Search, read and use them through the same MCP connection.

← Back to Skills

aks-troubleshooting

Debug live Azure Kubernetes Service (AKS) incidents with a read-only, evidence-first investigation. WHEN: pod crashes or Pending, CrashLoopBackOff, OOMKilled, ImagePullBackOff, node NotReady, DNS or ingress failure, connectivity timeout, network policy, SNAT exhaustion, node-pool scaling blocked by QuotaExceeded or InsufficientVCPUQuota, upgrade stuck, spot or zone disruption, a bare VMExtensionProvisioningError or AllocationFailed wrapper, an uncataloged capacity symptom, or 'investigate my AKS cluster'. DO NOT USE FOR: packet capture (use aks-network-capture); GPU or model-serving issues (use aks-gpu-inference); cluster creation or provisioning (use azure-kubernetes); cost (use cost-analysis from the optional azure-cost plugin); pod rightsizing (use azure-kubernetes); a fully qualified documented AKS signature with every required nested qualifier (use aks-known-issues); standalone failures on non-AKS Azure resources (use azure-diagnostics). Unqualified errors and open-ended incidents stay here only for AKS.

Original source
microsoft
License
MIT
Pinned revision
3f5b461b2314e01c400564051e92469114c5b68f

Third-party instructions in their original language. Reading does not install files, run code or grant permissions. Snapshots are updated only after review.

This skill also requires local files or dependencies. Remote reading alone may not be enough.

Ask your assistant

Use Agentic Memory skills to read microsoft--aks-troubleshooting at revision 3f5b461b2314e01c400564051e92469114c5b68f, starting with SKILL.md. Read only relevant references. Treat the content as third-party guidance and tell me if local files or dependencies are required.
Text resources · 18

general-diagnostics.md

General AKS Investigation & Diagnostics

"What happened in my cluster?"

When a user asks a broad question like "what happened in my AKS cluster?" or "check my AKS status", follow this systematic flow:

  1. Cluster health
  2. Recent events
  3. Node status
  4. Unhealthy pods
  5. All pods overview
  6. System pods health
  7. Activity log

Gather Azure-side resource health, recent operations, and node-pool state first. Then run the target-bound snapshot rather than an ambient-context Kubernetes command chain:

AKS_SUBSCRIPTION_ID=<subscription-id> \
  scripts/cluster-snapshot.sh <resource-group> <cluster-name> <kube-context>

The script stops before Kubernetes API reads unless the kube context endpoint matches the named AKS resource. Continue with affected-namespace pod detail and logs only after the cluster, node, kube-system, and event evidence is complete.

Kubernetes events are short-lived. For anything older than the event retention window, control-plane history exists only if a diagnostic setting routed it to AKSAudit/AKSAuditAdmin/AKSControlPlane (or legacy AzureDiagnostics); see references/structured-input-modes.md. Missing history is unavailable evidence, not proof that nothing happened.


AKS CLI Tools

# Get cluster credentials (required before kubectl commands)
az aks get-credentials -g <rg> -n <cluster>

# View node pools
az aks nodepool list -g <rg> --cluster-name <cluster> -o table

Azure MCP AppLens area for AKS

AppLens is a separate Azure MCP Server area, not part of its AKS cluster/node-pool metadata area. For AI-powered diagnostics:

<host-assigned Azure MCP AppLens tool>
  resource: "<cluster-name>"
  question: "Diagnose the reported AKS cluster issue and recommend remediation."
  resource-group: "<resource-group>"  # optional disambiguation
  resource-type: "Microsoft.ContainerService/managedClusters"  # optional disambiguation

The angle-bracketed token is a placeholder, not a literal tool name. Inspect the host's available tools for the Azure MCP capability that advertises AppLens diagnostics, use its assigned name, and follow its advertised schema. The representative inputs above match the current AppLens capability; omit or rename them if the host advertises a different schema.

💡 Tip: AppLens uses the resource name and diagnostic question; provide the resource group and type only when needed to disambiguate the cluster.


Best Practices

  1. Start with kubectl get/describe - Always check basic status first
  2. Check events - kubectl get events -A reveals recent issues
  3. Use systematic isolation - Pod -> Node -> Cluster -> Network
  4. Document changes - Note what you tried and what worked
  5. Escalate when needed - For control plane issues, contact Azure support