infra-ci-cd-docker（infra CI CD Docker 部署）

Docker Containerization Patterns

Quick Guide: Docker with BuildKit for containerizing Node.js/TypeScript applications. Multi-stage builds for minimal production images (1GB to under 100MB). Docker Compose v2 for development environments. BuildKit cache mounts for 10x faster dependency installs. Non-root users, health checks, and secret mounts for production security. Alpine for size, Debian slim for compatibility.

<critical_requirements>

CRITICAL: Before Using This Skill

All code must follow project conventions in CLAUDE.md (kebab-case, named exports, import ordering, import type, named constants)

(You MUST use multi-stage builds for production images - NEVER ship dev dependencies, TypeScript compiler, or source files in production)

(You MUST run containers as non-root user - NEVER run production containers as root (default))

(You MUST use CMD ["node", "server.js"] (exec form) - NEVER use npm start or shell form as CMD)

(You MUST use BuildKit secret mounts for sensitive data at build time - NEVER use ARG or ENV for secrets)

(You MUST copy package.json/lockfile BEFORE source code - NEVER COPY.. before npm ci (breaks layer cache))

</critical_requirements>

Examples

• Dockerfile Patterns - Multi-stage builds, Bun, monorepo, layer caching,.dockerignore, signal handling
• Docker Compose - Development environments, networking, volumes, healthchecks
• Production & CI/CD - Security hardening, secrets, CI/CD pipelines, vulnerability scanning
• Quick Reference - Dockerfile instructions, CLI commands, base image comparison

Auto-detection: Dockerfile, docker-compose, compose.yaml, Docker, container, multi-stage build, BuildKit,.dockerignore, Docker Compose, docker build, docker run, HEALTHCHECK, Docker Scout, docker init, containerize, container image, Docker network, Docker volume

When to use:

• Creating Dockerfiles for Node.js/TypeScript applications
• Setting up multi-stage builds to minimize production image size
• Configuring Docker Compose for local development environments
• Optimizing Docker layer caching and BuildKit cache mounts
• Implementing container security (non-root, secrets, read-only filesystem)
• Setting up health checks for container orchestration
• Building CI/CD pipelines that build and push Docker images
• Teams needing consistent development environments across OS platforms
• Microservice architectures requiring isolated services with dependencies

When NOT to use:

• Serverless deployments (AWS Lambda, Vercel Functions) that don't use containers
• Static site hosting (Netlify, Vercel, Cloudflare Pages) with no server runtime
• Simple scripts or CLI tools distributed via npm
• Local development without containerization requirements
• When added complexity outweighs the isolation benefit
• Kubernetes-specific orchestration patterns (use a Kubernetes skill)

Key patterns covered:

• Multi-stage Dockerfile for Node.js/TypeScript (builder pattern)
• Docker Compose v2 development environments
• BuildKit cache mounts and layer optimization
• Container security (non-root, secrets, capabilities, read-only)
• Health checks for production containers
• .dockerignore for build context optimization
• Volume mounts (named volumes, bind mounts, tmpfs)
• Docker networking (bridge, host, overlay)
• CI/CD integration (GitHub Actions build-push)
• Signal handling (tini for graceful shutdown)

Philosophy

Containers provide reproducible, isolated environments that eliminate "works on my machine" problems. Docker is the standard for packaging Node.js/TypeScript applications into portable, lightweight images.

Core principles:

1. Minimal production images - Ship only what the app needs to run (compiled JS, production deps, runtime)
2. Layer cache optimization - Structure Dockerfiles so unchanged layers are reused, making rebuilds fast
3. Security by default - Non-root users, no secrets in images, minimal attack surface
4. Development parity - Docker Compose mirrors production topology locally

Core Patterns

Pattern 1: Multi-Stage Dockerfile for Node.js/TypeScript

Multi-stage builds compile TypeScript in a builder stage, then copy only compiled JS and production dependencies into a minimal runtime image. This reduces images from 1GB+ to under 100MB.

A production Dockerfile uses three stages:

1. deps - Install production dependencies only
2. builder - Install all dependencies, compile TypeScript
3. runner - Copy compiled output and production deps into minimal image

# Stage 1: Production dependencies
FROM node:22-alpine AS deps
WORKDIR /app
COPY package.json package-lock.json ./
RUN --mount=type=cache,target=/root/.npm \
    npm ci --omit=dev --no-audit --no-fund

# Stage 2: Build TypeScript
FROM node:22-alpine AS builder
WORKDIR /app
COPY package.json package-lock.json ./
RUN --mount=type=cache,target=/root/.npm \
    npm ci --no-audit --no-fund
COPY tsconfig.json ./
COPY src/ ./src/
RUN npm run build

# Stage 3: Production runtime
FROM node:22-alpine AS runner
WORKDIR /app
RUN apk add --no-cache tini
ENV NODE_ENV=production
RUN addgroup --system --gid 1001 appgroup && \
    adduser --system --uid 1001 --ingroup appgroup appuser
COPY --from=deps --chown=appuser:appgroup /app/node_modules ./node_modules
COPY --from=builder --chown=appuser:appgroup /app/dist ./dist
COPY --chown=appuser:appgroup package.json ./
USER appuser
EXPOSE 3000
ENTRYPOINT ["/sbin/tini", "--"]
CMD ["node", "dist/server.js"]

Why good: Three-stage build separates concerns, BuildKit cache mount speeds up npm ci, non-root user, tini for signal handling, only production artifacts in final image

See examples/core.md for complete Dockerfiles including Bun and monorepo variants.

Pattern 2: Docker Compose for Development

Docker Compose v2 defines multi-container development environments. Use compose.yaml (not docker-compose.yml) with the docker compose command (no hyphen).

# compose.yaml
services:
  app:
    build:
      context: .
      target: builder
    volumes:
      - .:/app:cached
      - /app/node_modules
    depends_on:
      db:
        condition: service_healthy
  db:
    image: postgres:17-alpine
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U dev"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 30s
    volumes:
      - pgdata:/var/lib/postgresql/data
volumes:
  pgdata:

Why good: Health checks on dependencies with condition: service_healthy, bind mount with :cached for macOS perf, anonymous volume protects node_modules, named volume for database persistence

See examples/compose.md for full development environments with Redis, networking, and volume patterns.

Pattern 3: BuildKit Cache Mounts

BuildKit (default since Docker Engine 23+) provides cache mounts that persist package manager caches across builds, reducing install times by 10x or more.

# npm
RUN --mount=type=cache,target=/root/.npm \
    npm ci --no-audit --no-fund

# pnpm
RUN --mount=type=cache,target=/root/.local/share/pnpm/store \
    pnpm install --frozen-lockfile

Layer ordering rules: Pin base image versions > Copy dependency manifests > Install dependencies > Copy source code > Build/compile. This maximizes cache hits.

See examples/core.md for cache mount patterns for npm, Bun, pnpm, and apt.

Pattern 4: Container Security

Non-Root User

RUN addgroup --system --gid 1001 appgroup && \
    adduser --system --uid 1001 --ingroup appgroup appuser
COPY --chown=appuser:appgroup --from=builder /app/dist ./dist
USER appuser

Secret Mounts (Build Time)

# Secret is ephemeral - never persisted in image layers
RUN --mount=type=secret,id=npm_token,env=NPM_TOKEN \
    npm ci --no-audit --no-fund
# Build: docker build --secret id=npm_token,env=NPM_TOKEN .

Runtime Hardening

services:
  app:
    read_only: true
    tmpfs: [/tmp]
    security_opt: [no-new-privileges:true]
    cap_drop: [ALL]

See examples/production.md for complete production compose, secrets management, and CI/CD pipelines.

Pattern 5: Health Checks

Health checks enable orchestrators to detect unresponsive containers and restart them automatically.

HEALTHCHECK --interval=30s --timeout=10s --retries=3 --start-period=30s \
    CMD ["node", "-e", "fetch('http://localhost:3000/health').then(r => { if (!r.ok) process.exit(1) }).catch(() => process.exit(1))"]

Why good: Uses Node.js built-in fetch (no curl needed in Alpine), start-period allows app startup, checks dedicated health endpoint not just root path

See examples/production.md for health check endpoint implementation and Compose healthcheck patterns.

Pattern 6:.dockerignore

node_modules
dist
.git
.env
.env.*
!.env.example
Dockerfile*
compose.yaml
__tests__
*.test.ts
coverage
.github

Why good: Excludes node_modules (reinstalled deterministically),.env files (prevents secret leaks),.git (invalidates cache), keeps build context small

See examples/core.md for a comprehensive.dockerignore file.

Pattern 7: Signal Handling and Graceful Shutdown

Node.js as PID 1 does not handle SIGTERM/SIGINT correctly. Use tini as init system.

RUN apk add --no-cache tini
ENTRYPOINT ["/sbin/tini", "--"]
CMD ["node", "dist/server.js"]

Why good: Tini forwards signals to node process, handles zombie process reaping, ensures graceful shutdown on docker stop

See examples/core.md for complete TypeScript graceful shutdown implementation.

Performance Optimization

Image Size Reduction

Build Speed Optimization

CI/CD Build Cache

# GitHub Actions - Cache Docker layers
- uses: docker/setup-buildx-action@v3
- uses: docker/build-push-action@v7
  with:
    cache-from: type=gha
    cache-to: type=gha,mode=max

Why good: GitHub Actions cache (type=gha) persists layers across CI runs, mode=max caches all layers (not just final)

See examples/production.md for complete CI/CD pipeline examples.

<decision_framework>

Decision Framework

Base Image Selection

Which base image?
  |
  +-- Need smallest image? --> Alpine (node:22-alpine, ~135MB)
  |     +-- musl libc compatibility issues? --> Use Debian slim instead
  |
  +-- Need maximum compatibility? --> Debian slim (node:22-slim, ~200MB)
  |     +-- Native extensions work out of the box
  |
  +-- Need debugging tools? --> Full image (node:22, ~1GB)
  |     +-- Development only, never for production
  |
  +-- Maximum security? --> Distroless (gcr.io/distroless/nodejs22-debian12)
        +-- No shell, no package manager, smallest attack surface

Compose vs Dockerfile Targets

How to manage dev vs prod?
  |
  +-- Multi-stage Dockerfile with targets
  |     +-- `docker compose build --target builder` for dev
  |     +-- Final stage for production
  |
  +-- Separate compose files
        +-- compose.yaml (base)
        +-- compose.override.yaml (dev - auto-loaded)
        +-- compose.prod.yaml (production)

Volume Strategy

What data needs to persist?
  |
  +-- Source code (dev hot reload) --> Bind mount with :cached
  +-- Database files --> Named volume
  +-- Temporary/cache data --> tmpfs
  +-- Secrets at runtime --> Docker secrets or tmpfs
  +-- node_modules in dev --> Anonymous volume (protect from host)

</decision_framework>

Integration Guide

Docker ecosystem tools:

• Docker Scout: docker scout cves for vulnerability scanning in CLI and CI
• BuildKit: Default build engine since Docker Engine 23+, enables cache mounts, secret mounts, multi-platform builds

CI/CD integration:

Docker images integrate with any CI/CD pipeline. See examples/production.md for build-push patterns and vulnerability scanning workflows.

Container orchestration:

Production images work with any container orchestrator. Health checks (HEALTHCHECK instruction) and non-root users are universally required regardless of platform.

Replaces:

• VM-based development environments (Docker Compose provides lighter isolation)
• Manual server provisioning (Dockerfiles codify environment setup)
• Node version managers in containers (Dockerfile pins exact Node version via base image tag)

<red_flags>

RED FLAGS

High Priority Issues:

• Running production containers as root (default behavior - always add USER directive)
• Secrets in ARG/ENV/COPY (persist in image layers - use --mount=type=secret)
• Single-stage builds shipping dev dependencies and source code to production
• Using npm start as CMD (npm swallows SIGTERM - use CMD ["node", "dist/server.js"])
• No .dockerignore file (node_modules sent to daemon, secrets leaked into image)
• Using :latest tag for base images (non-deterministic builds)

Medium Priority Issues:

• COPY.. before npm ci (any file change invalidates dependency cache)
• No health check defined (orchestrator cannot detect unresponsive containers)
• docker-compose.yml with version: key (deprecated - use compose.yaml without version)
• depends_on without condition: service_healthy (container starts before dependency is ready)
• Using npm install instead of npm ci in Dockerfile (non-deterministic, slower)

Common Mistakes:

• Forgetting anonymous volume for node_modules in dev (/app/node_modules) causing host to overwrite container's modules
• Not using :cached on bind mounts on macOS (significant performance impact)
• Installing build tools (gcc, make, python) in the final stage instead of the builder stage
• Exposing database ports to host in production Compose (only needed for dev)

Gotchas & Edge Cases:

• Alpine uses musl libc - some npm packages with native C bindings (bcrypt, sharp, canvas) may fail; use npm rebuild or switch to Debian slim
• Node.js fetch() is available since Node 18+ (no need for curl in health checks on Alpine)
• BuildKit cache mounts require # syntax=docker/dockerfile:1 or Docker Engine 23+ with BuildKit enabled by default
• docker compose down -v removes named volumes (data loss) - use docker compose down without -v to preserve data
• Docker Desktop on macOS/Windows has different file system performance than Linux - bind mounts are slower
• COPY --chown is more efficient than COPY + RUN chown (one layer vs two)
• tini must be installed explicitly on Alpine (apk add --no-cache tini) - it is NOT included by default in node:alpine images
• The node_modules/.cache directory can grow large in development - add it to .dockerignore

</red_flags>

<critical_reminders>

CRITICAL REMINDERS

All code must follow project conventions in CLAUDE.md (kebab-case, named exports, import ordering, import type, named constants)

(You MUST use multi-stage builds for production images - NEVER ship dev dependencies, TypeScript compiler, or source files in production)

(You MUST run containers as non-root user - NEVER run production containers as root (default))

(You MUST use CMD ["node", "server.js"] (exec form) - NEVER use npm start or shell form as CMD)

(You MUST use BuildKit secret mounts for sensitive data at build time - NEVER use ARG or ENV for secrets)

(You MUST copy package.json/lockfile BEFORE source code - NEVER COPY.. before npm ci (breaks layer cache))

Failure to follow these rules will result in bloated images (1GB+), security vulnerabilities (root access, leaked secrets), broken graceful shutdown (lost requests on deploy), and slow CI builds (no layer caching).

</critical_reminders>