Azure AKS GitOps Boilerplate with Terraform, ArgoCD and GitLab CI/CD

Timeline & Info

Overview

Engineered a production-ready Azure Kubernetes Service boilerplate that combines Terraform-provisioned hub-spoke networking with pull-based GitOps via ArgoCD, delivering a fully reproducible platform deployable from a single git push. Designed the architecture around seven independent security layers - from WAF OWASP 3.2 at the perimeter to Calico network policies and zero secrets in Git - while integrating managed data services (Azure SQL, PostgreSQL, Redis) through private endpoints. Implemented centralized multi-cluster management with Rancher, external SSO via Keycloak / Entra ID, IoT MQTT messaging through an EMQX operator, and a complete observability stack with Prometheus, Grafana, and Loki backed by Azure Blob Storage.

Architecture

The platform is organized across two provisioning domains and a management layer:

• Hub VNet (optional shared firewall)
  • Azure Firewall with DNAT rules and FQDN-based egress filtering
  • Hub to Spoke VNet peering
• Spoke VNet - AKS cluster (Azure CNI + Calico, multi-zone)
  • System node pool: ArgoCD, Traefik, cert-manager, External Secrets Operator, EMQX Operator, Prometheus + Grafana + Loki
  • Workload node pool: EMQX Cluster, application deployments
  • Data services: Azure SQL, PostgreSQL Flexible Server, Azure Redis - all via private endpoints
• Management Cluster
  • Rancher for multi-cluster RBAC and developer access
  • GitLab Runners for CI/CD pipeline execution

Platform Overview

%%{init: {'theme': 'base', 'themeVariables': {'fontSize': '14px', 'fontFamily': 'arial', 'primaryColor': '#1e293b', 'primaryTextColor': '#ffffff', 'primaryBorderColor': '#475569', 'lineColor': '#94a3b8', 'secondaryColor': '#1e3a5f', 'secondaryTextColor': '#ffffff', 'secondaryBorderColor': '#475569', 'tertiaryColor': '#16213e', 'tertiaryTextColor': '#ffffff', 'tertiaryBorderColor': '#475569', 'nodeBorder': '#475569', 'nodeTextColor': '#ffffff', 'mainBkg': '#1e293b', 'clusterBkg': '#0f172a', 'clusterBorder': '#475569', 'titleColor': '#ffffff', 'edgeLabelBackground': '#1e293b', 'textColor': '#ffffff', 'background': '#0f172a'}}}%%
graph LR
    subgraph ENTRY["EXTERNAL TRAFFIC"]
        direction TB
        USER["Users<br/>HTTPS :443"]
        IOT["IoT Devices<br/>MQTTS :8883"]
        ADMIN["Platform Admins"]
    end

    subgraph AZURE["AZURE CLOUD"]
        direction TB

        subgraph INGRESS["INGRESS LAYER"]
            direction TB
            subgraph AGWSN["Application Gateway Subnet"]
                PIP["AGW Public IP"]
                AGW["WAF v2<br/>OWASP 3.2"]
                PIP --> AGW
            end
            subgraph HUBFW["Hub VNet - Shared Firewall - Optional"]
                FWPIP["Firewall Public IP"]
                FW["Azure Firewall<br/>DNAT + Egress FQDN Filter"]
                FWPIP --> FW
            end
        end

        subgraph SPOKE["SPOKE VNET - 10.1.0.0/16"]
            direction TB

            subgraph K8S["AKS CLUSTER - Azure CNI - Calico - Multi-Zone"]
                direction LR
                subgraph SYSPOOL["SYSTEM NODE POOL"]
                    direction TB
                    ARGO["ArgoCD"]
                    TFK["Traefik<br/>Internal LB - L4"]
                    CERTM["cert-manager"]
                    ESO["External Secrets"]
                    EMQXOP["EMQX Operator"]
                    PROM["Prometheus<br/>Grafana - Loki"]
                end
                subgraph WRKPOOL["WORKLOAD NODE POOL"]
                    direction TB
                    EMQX["EMQX Cluster<br/>MQTT Broker"]
                    APPS["Your Applications"]
                end
            end

            subgraph DATASVC["DATA SERVICES - Private Endpoints"]
                direction LR
                KV["Key Vault<br/>Secrets"]
                STOR["Storage Account<br/>Loki Logs"]
                SQL["Azure SQL<br/>Private Link"]
                PSQL["PostgreSQL<br/>Flexible Server"]
                REDIS["Azure Redis<br/>Private Link"]
            end
        end
    end

    subgraph MGMT["MANAGEMENT CLUSTER"]
        direction TB
        RANCHER["Rancher<br/>Multi-Cluster RBAC"]
        GITLAB_R["GitLab Runners<br/>CI/CD Execution"]
    end

    subgraph IDP["EXTERNAL IDENTITY PROVIDER"]
        KC["Keycloak / Entra ID<br/>SSO - OIDC - SAML"]
    end

    subgraph GIT["GIT REPOSITORY"]
        direction TB
        G1["hub-infrastructure/"]
        G2["infrastructure/"]
        G3["system/"]
        G4["application/"]
    end

    USER -->|":443"| PIP
    IOT -->|":8883"| FWPIP
    ADMIN --> RANCHER
    AGW -->|"L7"| K8S
    FW -->|"DNAT :8883"| TFK
    HUBFW <-.->|"VNet Peering"| SPOKE
    RANCHER -.->|"Manages"| K8S
    KC -.->|"OIDC"| PROM
    KC -.->|"OIDC"| RANCHER
    ARGO -.->|"GitOps Pull system/ & application/"| GIT
    ESO -.->|"Secrets"| KV
    PROM -.->|"Logs"| STOR
    APPS -.->|"Private Link"| SQL
    APPS -.->|"Private Link"| REDIS
    APPS -.->|"VNet"| PSQL

Traffic Flows

%%{init: {'theme': 'base', 'themeVariables': {'fontSize': '14px', 'fontFamily': 'arial', 'primaryColor': '#1e293b', 'primaryTextColor': '#ffffff', 'primaryBorderColor': '#475569', 'lineColor': '#94a3b8', 'secondaryColor': '#1e3a5f', 'secondaryTextColor': '#ffffff', 'secondaryBorderColor': '#475569', 'tertiaryColor': '#16213e', 'tertiaryTextColor': '#ffffff', 'tertiaryBorderColor': '#475569', 'nodeBorder': '#475569', 'nodeTextColor': '#ffffff', 'mainBkg': '#1e293b', 'clusterBkg': '#0f172a', 'clusterBorder': '#475569', 'titleColor': '#ffffff', 'edgeLabelBackground': '#1e293b', 'textColor': '#ffffff', 'background': '#0f172a'}}}%%
graph LR
    subgraph HTTPS["HTTPS - Web / API / SSO"]
        direction LR
        A1["User"] -->|":443"| A2["AGW Public IP"] --> A3["WAF v2"] -->|"L7"| A4["App / Grafana"]
    end

    subgraph MQTTS["MQTTS - IoT Devices"]
        direction LR
        B1["Device"] -->|":8883"| B2["FW Public IP"] --> B3["DNAT"] -->|"L4"| B4["Traefik"] --> B5["EMQX Broker"]
    end

    subgraph DATA["DATA - Application to Database"]
        direction LR
        D1["App Pod"] -->|"Private Link"| D2["MSSQL :1433"]
        D1 -->|"VNet"| D3["PostgreSQL :5432"]
        D1 -->|"Private Link"| D4["Redis :6380"]
    end

    subgraph EGRESS["EGRESS - Outbound via Hub-Spoke"]
        direction LR
        C1["AKS Pod"] --> C2["UDR"] --> C3["FW FQDN Filter"] --> C4["Internet"]
    end

Implementation

Infrastructure as Code

Provisioned the complete Azure networking topology with Terraform - Spoke VNet, subnets, NSG rules, Application Gateway WAF v2 with 10 security headers, AKS cluster with system and workload node pools, Key Vault with RBAC, private endpoints for all data services, and managed identities for workload authentication. Structured the IaC into two independent Terraform roots (hub-infrastructure/ and infrastructure/) to allow hub-spoke deployments to be adopted incrementally, keeping single-cluster setups simpler while enabling full enterprise network topology when needed.

CI/CD Pipeline

Implemented a GitLab CI/CD pipeline with five ordered stages: secret-detection - validate - plan - apply (manual gate) - destroy (manual gate). Integrated Gitleaks as the first pipeline stage to block secrets from reaching the Terraform plan, ensuring the repo remains safe to share and audit at any point in the lifecycle.

GitOps with ArgoCD

Deployed the App of Apps pattern where a single root ArgoCD application bootstraps all 10 platform applications from the system/argocd/applications/ directory. This approach means every platform service - Traefik, cert-manager, External Secrets, monitoring, Loki, EMQX Operator - is self-managed through Git, with ArgoCD continuously reconciling live cluster state against the repository.

Secrets Management

Implemented zero secrets in Git by routing all sensitive values through External Secrets Operator backed by Azure Key Vault. A ClusterSecretStore authenticates to Key Vault using a managed identity, and individual ExternalSecret resources in system/ and application/ pull values at runtime - covering storage keys, Grafana admin credentials, Git repository tokens, and application secrets.

Networking & Ingress

Configured dual ingress paths:

• HTTPS - Application Gateway WAF v2 (AGIC) for L7 traffic with OWASP 3.2 Prevention mode and TLS termination
• MQTTS - Azure Firewall DNAT on port 8883 forwarding to Traefik internal load balancer, which routes to the EMQX cluster

cert-manager handles automated TLS lifecycle for both paths using ClusterIssuers targeting Let’s Encrypt.

System Components

Security Design

Implemented seven independent defense layers with centralized identity:

Key Results

Tech Stack

Skills Demonstrated

Infrastructure as Code, Azure Cloud Architecture, Kubernetes Platform Engineering, GitOps Design, CI/CD Pipeline Automation, Secret Management, Network Security, WAF Configuration, Multi-Cluster Management, Identity Federation, Observability Engineering, Declarative Configuration, Technical Documentation