Planifier et configurer un cluster AKS prêt pour la production

---
name: azure-kubernetes
license: MIT
metadata:
  author: Microsoft
  version: "1.2.5"
description: "Plan, create, and configure production-ready Azure Kubernetes Service (AKS) clusters. Covers Day-0 checklist, SKU selection (Automatic vs Standard), networking options (private API server, Azure CNI Overlay, egress configuration), security, and operations (autoscaling, upgrade strategy, cost analysis). WHEN: create AKS environment, provision AKS, enable AKS observability, design AKS networking, choose AKS SKU, secure AKS, optimize AKS, AKS spot nodes, AKS cluster-autoscaler, rightsize AKS pod, pod rightsizing, over-provisioned AKS pod, pod resource requests and limits, Vertical Pod Autoscaler, VPA recommendations."
---

# Azure Kubernetes Service

> **AUTHORITATIVE GUIDANCE — MANDATORY COMPLIANCE**
>
> This skill produces a **recommended AKS cluster configuration** based on user requirements, distinguishing **Day-0 decisions** (networking, API server — hard to change later) from **Day-1 features** (can enable post-creation). See [CLI reference](./references/cli-reference.md) for commands.

## Quick Reference
| Property | Value |
|----------|-------|
| Best for | AKS cluster planning and Day-0 decisions |
| MCP Tools | `mcp_azure_mcp_aks` |
| CLI | `az aks create`, `az aks show`, `kubectl get`, `kubectl describe` |
| Related skills | azure-kubernetes-app-deploy (deploy an app to an existing cluster), azure-diagnostics (baseline AKS troubleshooting), azure-validate (readiness checks), azure-kubernetes-automatic-readiness (migrate existing cluster to AKS Automatic), optional `aks-skills` operations add-on (focused incident, advanced cost-efficiency, exact-signature, packet-capture, and GPU Day-2 workflows) |

## When to Use This Skill
Activate this skill when user wants to:
- Create a new AKS cluster
- Plan AKS cluster configuration for production workloads
- Design AKS networking (API server access, pod IP model, egress)
- Set up AKS identity and secrets management
- Configure AKS governance (Azure Policy, Deployment Safeguards)
- Enable AKS observability (Container Insights, Managed Prometheus, Grafana)
- Define AKS upgrade and patching strategy
- Understand AKS Automatic vs Standard SKU differences
- Get a Day-0 checklist for AKS cluster setup and configuration

> **Deploying an application to an existing cluster?** This skill provisions and
> configures the *cluster*. To containerize an app and deploy it to a cluster
> that already exists (Dockerfile + manifests + Deployment Safeguards), use the
> `azure-kubernetes-app-deploy` sub-skill instead.

> **Deeper operational task?** Keep Day-0 planning and basic operations in this
> base skill. Load [optional-aks-operations.md](./references/optional-aks-operations.md)
> only when a focused optional AKS operational skill could improve the current
> task.

## Rules
1. Start with the user's requirements for provisioning compute, networking, security, and other settings.
2. Use the `azure` MCP server and select `mcp_azure_mcp_aks` first to discover the exact AKS-specific MCP tools surfaced by the client. Choose the smallest discovered AKS tool that fits the task, and fall back to Azure CLI (`az aks`) only when the needed functionality is not exposed through the AKS MCP surface.
3. Determine if AKS Automatic or Standard SKU is more appropriate based on the user's need for control vs convenience. Default to AKS Automatic unless specific customizations are required.
4. Document decisions and rationale for cluster configuration choices, especially for Day-0 decisions that are hard to change later (networking, API server access).


## Required Inputs (Ask only what’s needed)
If the user is unsure, use safe defaults.
- AKS environment type: dev/test or production
- Region(s), availability zones, preferred node VM sizes
- Expected scale (node/cluster count, workload size)
- Networking requirements (API server access, pod IP model, ingress/egress control)
- Security and identity requirements, including image registry
- Upgrade and observability preferences
- Cost constraints

## Workflow

### 1. Cluster Type
- **AKS Automatic** (default): Best for most production workloads, provides a curated experience with pre-configured best practices for security, reliability, and performance. Use unless you have specific custom requirements for networking, autoscaling, or node pool configurations not supported by Node Auto-Provisioning (NAP).
- **AKS Standard**: Use if you need full control over environment configuration, which requires additional overhead to set up and manage.

### 2. Networking (Pod IP, Egress, Ingress, Dataplane)

**Pod IP Model** (Key Day-0 decision):
- **Azure CNI Overlay** (recommended): pod IPs from private overlay range, not VNet-routable, scales to large environments and good for most workloads
- **Azure CNI (VNet-routable)**: pod IPs directly from VNet (pod subnet or node subnet), use when pods must be directly addressable from VNet or on-prem
  - Docs: https://learn.microsoft.com/azure/aks/azure-cni-overlay

**Dataplane & Network Policy**:
- **Azure CNI powered by Cilium** (recommended): eBPF-based for high-performance packet processing, network policies, and observability

**Egress**:
- **Static Egress Gateway** for stable, predictable outbound IPs
- For restricted egress: UDR + Azure Firewall or NVA

**Ingress**:
- **App Routing addon with Gateway API** — recommended default for HTTP/HTTPS workloads
- **Istio service mesh with Gateway API** - for advanced traffic management, mTLS, canary releases
- **Application Gateway for Containers** — for L7 load balancing with WAF integration

**DNS**:
- Enable **LocalDNS** on all node pools for reliable, performant DNS resolution

### 3. Security
- Use **Microsoft Entra ID** everywhere (control plane, Workload Identity for pods, node access). Avoid static credentials.
- Azure Key Vault via **Secrets Store CSI Driver** for secrets
- Enable **Azure Policy** + **Deployment Safeguards**
- Enable **Encryption at rest** for etcd/API server; **in-transit** for node-to-node
- Allow only signed, policy-approved images (Azure Policy + Ratify), prefer **Azure Container Registry**
- **Isolation**: Use namespaces, network policies, scoped logging

### 4. Observability
- Use Managed Prometheus and Container Insights with Grafana for AKS observability (logs + metrics).
- Enable Diagnostic Settings to collect control plane logs and audit logs in a Log Analytics workspace for security monitoring and troubleshooting.
- For other monitoring and troubleshooting tools, use features like the Agentic CLI for AKS, Application Insights, Resource Health Center, AppLens detectors, and Azure Advisors.

### 5. Upgrades & Patching
- Configure **Maintenance Windows** for controlled upgrade timing
- Enable **auto-upgrades** for control plane and node OS to stay up-to-date with security patches and Kubernetes versions
- Consider **LTS versions** for enterprise stability (2-year support) by upgrading your AKS environment to the Premium tier
- **Fleet upgrades**: Use **AKS Fleet Manager** for staged rollout across test to production environments

### 6. Performance
- Use **Ephemeral OS disks** (`--node-osdisk-type Ephemeral`) for faster node startup
- Select **Azure Linux** as node OS (smaller footprint, faster boot)
- Enable **KEDA** for event-driven autoscaling beyond HPA

### 7. Node Pools & Compute
- **Dedicated system node pool**: At least 2 nodes, tainted for system workloads only (`CriticalAddonsOnly`)
- **Node Auto Provisioning (NAP)** is enabled per cluster and can reduce cost
  when dynamic capacity fits workload demand; savings are not guaranteed.
  Enable NAP for new clusters. For existing clusters, use
  [optional AKS operations](./references/optional-aks-operations.md) to invoke
  `aks-cost-optimization` when available for NAP readiness. Otherwise, continue with
  [base NAP readiness](./references/azure-aks-nap-readiness.md).
- Use **latest generation SKUs (v5/v6)** for host-level optimizations
- **Avoid B-series VMs** — burstable SKUs cause performance/reliability issues
- Use SKUs with **at least 4 vCPUs** for production workloads
- Set **topology spread constraints** to distribute pods across hosts/zones per SLO

### 8. Reliability
- Deploy across **3 Availability Zones** (`--zones 1 2 3`)
- Use **Standard tier** for zone-redundant control plane + 99.95% SLA for API server availability
- Enable **Microsoft Defender for Containers** for runtime protection
- Configure **PodDisruptionBudgets** for all production workloads
- Use **topology spread constraints** to ensure pod distribution across failure domains

### 9. Cost Controls
- Use **Spot node pools** for batch/interruptible workloads (up to 90% savings)
- **Stop/Start** dev/test clusters: `az aks stop/start`

#### Focused AKS Cost Workflows

For a focused request, load only the relevant reference file.

| Scenario | Trigger Keywords | Reference |
|----------|-----------------|-----------|
| Pod Rightsizing | over-provisioned pods, CPU requests, memory requests, rightsize workloads | [azure-aks-rightsizing.md](./references/azure-aks-rightsizing.md) |
| VPA Setup | vertical pod autoscaler, VPA recommendations, VPA enable | [azure-aks-vpa.md](./references/azure-aks-vpa.md) |
| Cluster Autoscaler | idle nodes, CAS off, enable autoscaler, scale-down profile, node utilization | [azure-aks-autoscaler.md](./references/azure-aks-autoscaler.md) |
| Spot Node Pools | Spot VMs, Spot nodes, batch workloads, cheaper nodes | [azure-aks-spot.md](./references/azure-aks-spot.md) |
| Advanced AKS Cost Efficiency | NAP readiness, node-pool bin packing, blocked scale-down, Kubernetes storage cost | [optional AKS operations](./references/optional-aks-operations.md) |

> **Disambiguation:** If a prompt matches multiple rows (e.g., "cheaper nodes" could suggest both Spot and autoscaler), prefer the most specific match. If ambiguous, ask the user to clarify their intent before loading a reference file.

#### Comprehensive AKS Cost Review

When the user requests an overall AKS cost assessment rather than one focused
workflow, this skill owns triage and coordinates one unified review:

1. Confirm the target subscription, resource group, and AKS cluster. If `cost-analysis` is available, collect one billed-cost baseline for that scope and reuse it throughout the review.
2. Evaluate the applicable rightsizing, VPA, Cluster Autoscaler, and Spot workflows.
3. Invoke `aks-cost-optimization` when available for NAP readiness, node-pool bin packing, node-scaling bottlenecks, or Kubernetes-to-Azure storage optimization.
4. Check for stable non-Spot demand. If present, invoke `cost-optimization` when available for commitment analysis. If the capability or evidence is unavailable, report commitment analysis as a gap. Do not recommend a purchase from this review alone.

Investigate applicable levers and return one ranked report. Rank by comparable
estimated cost impact when available; keep unknown monetary impact unknown
rather than zero. Do not present one narrow analysis as the complete review.
Mark unavailable capabilities or evidence as gaps.

| Rank | Lever | Evidence | Estimated cost impact and basis | Risk | Approval needed | Confidence |
|---|---|---|---|---|---|---|

## Guardrails / Safety
- Do not request or output secrets (tokens, keys).
- Do not ask the user to paste subscription IDs. Discover subscription and resource scope via MCP tools (e.g., list subscriptions, list resource groups) or `az account show` / `az account list` so the agent can resolve context without exposing identifiers.
- If requirements are ambiguous for day-0 critical decisions, ask the user clarifying questions. For day-1 enabled features, propose 2–3 safe options with tradeoffs and choose a conservative default.
- Do not promise zero downtime; advise workload safeguards (PDBs, probes, replicas) and staged upgrades along with best practices for reliability and performance.

## MCP Tools
| Tool | Purpose | Key Parameters |
|------|---------|----------------|
| `mcp_azure_mcp_aks` | AKS MCP entry point used to discover the exact AKS-specific tools exposed by the client | Discover the callable AKS tool first, then use that tool's parameters |

## Error Handling
| Error / Symptom | Likely Cause | Remediation |
|-----------------|--------------|-------------|
| MCP tool call fails or times out | Invalid credentials, subscription, or AKS context | Verify `az login`, confirm the active subscription context with `az account show`, and check the target resource group without echoing subscription identifiers back to the user |
| Quota exceeded | Regional vCPU or resource limits | Request quota increase or select different region/VM SKU |
| Networking conflict (IP exhaustion) | Pod subnet too small for overlay/CNI | Re-plan IP ranges; may require cluster recreation (Day-0) |
| Workload Identity not working | Missing OIDC issuer or federated credential | Enable `--enable-oidc-issuer --enable-workload-identity`, configure federated identity |