nestjs-deployment
$
npx mdskill add HoangNguyen0403/agent-skills-standard/nestjs-deployment1. **Write multi-stage Dockerfile** — Build stage installs devDeps and runs `nest build`; run stage copies only `dist` and pruned `node_modules`. 2. **Set non-root user** — Add `USER node` to Dockerfile. 3. **Tune memory** — Set `--max-old-space-size` to ~75% of container memory limit. 4. **Enable shutdown hooks** — Call `app.enableShutdownHooks()` in `main.ts`. 5. **Add K8s pre-stop** — Configure 5-10s sleep pre-stop hook for LB draining.
SKILL.md
.github/skills/nestjs-deploymentView on GitHub ↗
---
name: nestjs-deployment
description: Containerize NestJS apps with multi-stage Docker builds, tune Node.js memory, and implement graceful shutdown hooks. Use when writing Dockerfiles, configuring K8s deployments, or adding shutdown hooks for NestJS.
metadata:
triggers:
files:
- 'k8s/**'
- 'helm/**'
keywords:
- Dockerfile
- max-old-space-size
- shutdown hooks
---
# Deployment & Ops Standards
## **Priority: P1 (OPERATIONAL)**
## Workflow: Containerize NestJS App
1. **Write multi-stage Dockerfile** — Build stage installs devDeps and runs `nest build`; run stage copies only `dist` and pruned `node_modules`.
2. **Set non-root user** — Add `USER node` to Dockerfile.
3. **Tune memory** — Set `--max-old-space-size` to ~75% of container memory limit.
4. **Enable shutdown hooks** — Call `app.enableShutdownHooks()` in `main.ts`.
5. **Add K8s pre-stop** — Configure 5-10s sleep pre-stop hook for LB draining.
## Dockerfile Example
See [implementation examples](references/example.md)
## Runtime Tuning (Node.js)
- **Memory Config**: Container memory != Node memory.
- **Rule**: Explicitly set Max Old Space.
- **Command**: `node --max-old-space-size=XXX dist/main`
- **Calculation**: Set to ~75-80% of Kubernetes Limit. (Limit: 1GB -> OldSpace: 800MB).
- **Graceful Shutdown**:
- **Signal**: Listen to `SIGTERM`.
- **NestJS**: `app.enableShutdownHooks()` mandatory.
- **Sleep**: Add "Pre-Stop" sleep in K8s (5-10s) to allow Load Balancer to drain connections before Node process stops accepting traffic.
## Init Patterns
- **Database Migrations**:
- **Anti-Pattern**: Running migration in `main.ts` on startup.
- **Pro Pattern**: Use **Init Container** in Kubernetes that runs `npm run typeorm:migration:run` before app container starts.
## Environment Variables & CI/CD
- **CI/CD Pipelines (GitHub, GitLab, Azure, etc.)**:
- If you modify `src/config/env.validation.ts` to add new environment variable, you **MUST** map it explicitly in your deployment pipeline/infrastructure-as-code.
- **Platform Context**:
- **Cloud Run/ECS**: Variables must explicitly passed in service definition.
- **Kubernetes**: New variables must added to `Deployment` manifest or `ConfigMap`/`Secret`.
- **Lambda/Serverless**: Must added to `serverless.yml` or provider console.
- **Fundamental Rule**: Application code configuration changes "breaking changes" for infrastructure layer. Never assume environment inheritance.
## Anti-Patterns
- **No migrations in main.ts**: Use K8s Init Containers or pre-deploy CI steps for migration runs.
- **No root user in Docker**: Always add `USER node` to Dockerfile; running as root security risk.
- **No unbounded Node heap**: Set `--max-old-space-size` to ~75% of container memory limit.