Cloud Deploy Blueprint

Overview

This skill captures the complete knowledge for deploying a multi-service application to cloud Kubernetes, based on battle-tested learnings from deploying TaskFlow (5 microservices) to Azure AKS.

When to Use

• Deploying to AKS, GKE, or DOKS
• Setting up CI/CD with GitHub Actions
• Integrating managed services (Neon PostgreSQL, Upstash Redis)
• Configuring ingress with SSL certificates
• Handling Next.js NEXTPUBLIC* variables in Docker/K8s

Architecture Pattern

                         INTERNET
                             │
                             ▼
                    ┌─────────────────┐
                    │  Load Balancer  │  (Single Public IP)
                    └────────┬────────┘
                             │
                    ┌────────▼────────┐
                    │ Ingress (Traefik│  Routes by subdomain
                    │   or nginx)     │
                    └────────┬────────┘
                             │
        ┌────────────────────┼────────────────────┐
        │                    │                    │
        ▼                    ▼                    ▼
  ┌──────────┐        ┌──────────┐        ┌──────────┐
  │   Web    │        │   SSO    │        │   MCP    │
  │ (PUBLIC) │        │ (PUBLIC) │        │ (PUBLIC) │
  └────┬─────┘        └────┬─────┘        └────┬─────┘
       │                   │                   │
       │              ┌────▼─────┐             │
       └──────────────►   API    ◄─────────────┘
                      │(INTERNAL)│
                      └────┬─────┘
                           │
              ┌────────────┴────────────┐
              ▼                         ▼
      ┌─────────────┐           ┌─────────────┐
      │    Neon     │           │   Upstash   │
      │ (Postgres)  │           │   (Redis)   │
      │  EXTERNAL   │           │  EXTERNAL   │
      └─────────────┘           └─────────────┘

Critical Concept: Build-Time vs Runtime Variables

The Problem

Next.js NEXTPUBLIC* variables are embedded at build time, not runtime. This means:

# WRONG: Setting NEXT_PUBLIC_* at runtime does NOTHING
ENV NEXT_PUBLIC_API_URL=https://api.example.com

# RIGHT: Must be set as build ARG
ARG NEXT_PUBLIC_API_URL=https://api.example.com
ENV NEXT_PUBLIC_API_URL=$NEXT_PUBLIC_API_URL

The Solution

1. In Dockerfile: Use ARG for NEXTPUBLIC* variables
2. In CI/CD: Pass --build-arg with domain-specific values
3. In values.yaml: These are NOT runtime configurable

Build-Time Variables (Next.js)

Runtime Variables (ConfigMaps/Secrets)

Internal K8s Service Names

Services communicate via K8s service names, NOT public URLs:

# CORRECT - Internal communication
SSO_URL: http://sso-platform:3001
API_URL: http://taskflow-api:8000

# WRONG - Don't use public URLs for internal traffic
SSO_URL: https://sso.example.com

GitHub Actions CI/CD Pattern

Selective Builds with Path Filters

jobs:
  changes:
    runs-on: ubuntu-latest
    outputs:
      api: ${{ steps.filter.outputs.api }}
      web: ${{ steps.filter.outputs.web }}
    steps:
      - uses: dorny/paths-filter@v3
        id: filter
        with:
          filters: |
            api:
              - 'apps/api/**'
            web:
              - 'apps/web/**'

  build-api:
    needs: changes
    if: needs.changes.outputs.api == 'true' || github.event_name == 'workflow_dispatch'

Next.js Build Args Pattern

- name: Build and push (web)
  uses: docker/build-push-action@v5
  with:
    build-args: |
      NEXT_PUBLIC_SSO_URL=https://sso.${{ vars.DOMAIN }}
      NEXT_PUBLIC_API_URL=https://api.${{ vars.DOMAIN }}
      NEXT_PUBLIC_APP_URL=https://${{ vars.DOMAIN }}

GitHub Secrets & Variables

Secrets (Sensitive)

NEON_SSO_DATABASE_URL
NEON_API_DATABASE_URL
NEON_CHATKIT_DATABASE_URL
NEON_NOTIFICATION_DATABASE_URL
UPSTASH_REDIS_HOST
UPSTASH_REDIS_PASSWORD
REDIS_URL
REDIS_TOKEN
BETTER_AUTH_SECRET
OPENAI_API_KEY
SMTP_USER
SMTP_PASSWORD
AZURE_CREDENTIALS (or GCP_CREDENTIALS)

Variables (Non-sensitive)

DOMAIN=example.com
CLOUD_PROVIDER=azure
AZURE_RESOURCE_GROUP=myapp-rg
AZURE_CLUSTER_NAME=myapp-cluster
INGRESS_CLASS=traefik

Helm Values Pattern

values-cloud.yaml (Committed, Non-sensitive defaults)

global:
  domain: ""  # Set via --set
  namespace: taskflow
  imagePullPolicy: Always

managedServices:
  neon:
    enabled: true
    # Connection strings injected via --set from secrets
  upstash:
    enabled: true
    # Credentials injected via --set from secrets

sso:
  enabled: true
  name: sso-platform
  postgresql:
    enabled: false  # Using Neon
  env:
    NODE_ENV: production
    BETTER_AUTH_URL: ""  # Set via --set

Helm --set Pattern

helm upgrade --install taskflow ./infrastructure/helm/taskflow \
  --values values-cloud.yaml \
  --set global.imageRegistry="ghcr.io/owner/repo" \
  --set global.imageTag="${{ github.sha }}" \
  --set "managedServices.neon.ssoDatabase=${{ secrets.NEON_SSO_DATABASE_URL }}" \
  --set "sso.env.BETTER_AUTH_SECRET=${{ secrets.BETTER_AUTH_SECRET }}"

CRITICAL: CPU Architecture Check

BEFORE ANY DEPLOYMENT, check your cluster's node architecture:

kubectl get nodes -o jsonpath='{.items[*].status.nodeInfo.architecture}'

• amd64 → Use platforms: linux/amd64
• arm64 → Use platforms: linux/arm64

ARM64 is increasingly common (Azure, AWS Graviton, Apple Silicon dev). Don't assume amd64!

Docker Build for Correct Architecture

- uses: docker/build-push-action@v5
  with:
    platforms: linux/arm64      # MATCH YOUR CLUSTER!
    provenance: false           # Avoid manifest issues
    no-cache: true              # When debugging

Why provenance: false? Buildx attestation creates complex manifest lists that can cause "no match for platform" errors. Disable for simple, reliable images.

Common Gotchas (Battle-Tested)

1. Logout Redirect to 0.0.0.0

Problem: request.url in K8s returns container bind address Solution: Use NEXTPUBLICAPP_URL env var for redirects

// WRONG
const response = NextResponse.redirect(new URL("/", request.url));

// RIGHT
const APP_URL = process.env.NEXT_PUBLIC_APP_URL || "http://localhost:3000";
const response = NextResponse.redirect(new URL("/", APP_URL));

2. Email Verification Redirect to localhost

Problem: Missing NEXTPUBLICCONTINUE_URL in SSO Dockerfile Solution: Add to Dockerfile and CD pipeline:

ARG NEXT_PUBLIC_CONTINUE_URL=http://localhost:3000
ENV NEXT_PUBLIC_CONTINUE_URL=$NEXT_PUBLIC_CONTINUE_URL

3. Browser Making Requests to localhost

Problem: NEXTPUBLIC not passed as build arg Solution: Check ALL NEXTPUBLIC variables systematically:

grep -r "NEXT_PUBLIC_" apps/web/src --include="*.ts" --include="*.tsx" | \
  grep -oE "NEXT_PUBLIC_[A-Z_]+" | sort -u

4. Hardcoded Sensitive Data

Problem: Email/passwords hardcoded in values files Solution: Use --set from GitHub Secrets for ALL sensitive data

5. Missing Database Sections in values.yaml

Problem: Helm templates expect database.host, postgresql.name etc. Solution: Include empty/default sections even for managed services:

postgresql:
  enabled: false
  name: sso-platform-postgres

database:
  host: ""
  port: "5432"
  name: taskflow_sso
  user: postgres

SSL Certificate Pattern (cert-manager)

apiVersion: cert-manager.io/v1
kind: ClusterIssuer
metadata:
  name: letsencrypt-prod
spec:
  acme:
    server: https://acme-v02.api.letsencrypt.org/directory
    email: [email protected]
    privateKeySecretRef:
      name: letsencrypt-prod
    solvers:
    - http01:
        ingress:
          class: traefik

Ingress Annotations for TLS

annotations:
  cert-manager.io/cluster-issuer: letsencrypt-prod
  traefik.ingress.kubernetes.io/router.tls: "true"

Pre-Deployment Checklist

Code Changes

• All NEXTPUBLIC* vars documented and in Dockerfiles
• Redirect URLs use env vars, not request.url
• No hardcoded localhost in production code paths

Dockerfiles

• All NEXTPUBLIC* as ARG and ENV
• Multi-stage build for slim production image
• Health check endpoint configured

CI/CD Pipeline

• Build args for Next.js apps
• Path filters for selective builds
• All secrets listed and documented
• Helm --set for all sensitive values

Helm Chart

• values-cloud.yaml has all required sections
• No sensitive data in committed files
• Internal service names for inter-service communication
• Ingress configured with correct class

GitHub Setup

• All secrets created in repository settings
• All variables created in repository settings
• Azure/GCP credentials configured

Related Skills

• aks-deployment-troubleshooter - Debug ImagePullBackOff, CrashLoopBackOff, architecture issues
• containerize-apps - Dockerization patterns
• helm-charts - Helm chart structure
• kubernetes-essentials - K8s fundamentals
• better-auth-sso - SSO integration

Related Agents

• impact-analyzer-agent - Pre-containerization analysis