cicd-pipeline-architecture

# CI/CD Pipeline Architecture

## Overview

**Design CI/CD pipelines with deployment verification, rollback capabilities, and zero-downtime strategies from day one.**

**Core principle:** "Deploy to production" is not a single step - it's a sequence of gates, health checks, gradual rollouts, and automated rollback triggers. Skipping these "for speed" causes production incidents.

## When to Use

Use this skill when:
- **Setting up new CI/CD pipelines** (before writing workflow files)
- **Experiencing deployment failures** in production
- **CI feedback loops are too slow** (tests taking too long)
- **No confidence in deployments** (fear of breaking production)
- **Manual rollbacks required** after bad deploys
- **Downtime during deployments** is acceptable (it shouldn't be)
- **Migrations cause production issues**

**Do NOT skip this for:**
- "Quick MVP" or "demo" pipelines (these become production)
- "Simple" applications (complexity comes from deployment, not code)
- "We'll improve it later" (later never comes, incidents do)

## Core Pipeline Architecture

### Mandatory Pipeline Stages

**Every production pipeline MUST include:**

```
1. Build → 2. Test → 3. Deploy to Staging → 4. Verify Staging → 5. Deploy to Production → 6. Verify Production → 7. Monitor
```

**Missing any stage = production incidents waiting to happen.**

### 1. Build Stage

**Purpose:** Compile, package, create artifacts

```yaml
build:
  - Compile code (if applicable)
  - Run linters and formatters
  - Build container image
  - Tag with commit SHA (NOT "latest")
  - Push to registry
  - Create immutable artifact
```

**Key principle:** Build once, deploy everywhere. Same artifact to staging and production.

### 2. Test Stage

**Test Pyramid in CI:**

```
       /\
      /E2\      ← Few, critical paths only (5-10 tests)
     /----\
    / Intg \    ← API contracts, DB integration (50-100 tests)
   /--------\
  /   Unit   \  ← Fast, isolated, thorough (100s-1000s)
 /____________\
```

**Optimizat