Capstone: Implementation & Deployment

The CI Pipeline

Every push runs the same gates a broken merge would skip:

name: ci
on: [push, pull_request]
jobs:
  verify:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with: { node-version: 20, cache: npm }
      - run: npm ci
      - run: npm run lint && npm run typecheck
      - run: npm test -- --coverage
      - run: npm run build

Order matters: cheapest feedback first (lint, tsc --noEmit), then tests, then build. Use npm ci — it installs from the lockfile exactly — and cache the package manager's store so cold runs stay fast. Failing the job on any step is what makes main trustworthy.

Build, Containerize, Release

Ship an artifact, not a checkout. A multi-stage Dockerfile keeps dev dependencies out of production:

FROM node:20-slim AS build
WORKDIR /app
COPY package*.json ./
RUN npm ci
COPY . .
RUN npm run build

FROM node:20-slim
WORKDIR /app
ENV NODE_ENV=production
COPY package*.json ./
RUN npm ci --omit=dev
COPY --from=build /app/dist ./dist
CMD ["node", "dist/server.js"]

Deployment runs migrations before switching traffic, then health checks the new version. Tag images by git SHA (app:4f9c2ab) so rollback is retagging to the previous SHA, not rebuilding. Configuration comes from the platform's environment; secrets never enter the image layer.

Production Pitfalls

Secrets in the repo. A leaked key in git history is compromised even after deletion — use platform secret stores, rotate on exposure, and keep .env out of images with .dockerignore.

No health endpoint. Orchestrators route traffic only when /healthz answers 200; report dependency status there (database ping) so half-up instances drain.

Migrations as deploy steps. Running migrate deploy in a pre-deploy job keeps old and new code compatible with the schema — prefer additive migrations so rollback does not need a schema downgrade.

Silent config drift. Validate env with a typed schema at boot and fail fast; missing PORT should crash at startup, not at first request.

Interview: describe your pipeline stages, the rollback path, and where a migration runs relative to the traffic switch.