Test Environments

local         Developer's machine. Fast iteration. No SLAs.
              Risk: "works on my machine" — config and service versions drift.

dev/ci        CI environment. Every PR. Automated tests run here.
              Should be closest to staging in config. Ephemeral where possible.

staging       Production mirror. Manual QA and integration testing.
              Same infrastructure class as prod, but scaled down.
              Shared between team — coordinate deployments.

production    Live. Real users. No testing except smoke/canary.

# GitHub Actions — deploy ephemeral environment per PR
name: PR Environment

on:
  pull_request:
    types: [opened, synchronize, reopened]

jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
    - name: Deploy to ephemeral env
      run: |
        # Build and push Docker image
        docker build -t myregistry/myapp:pr-${{ github.event.number }} .
        docker push myregistry/myapp:pr-${{ github.event.number }}

        # Deploy to Kubernetes with PR-specific namespace
        kubectl create namespace pr-${{ github.event.number }} --dry-run=client -o yaml | kubectl apply -f -
        helm upgrade --install myapp-pr-${{ github.event.number }} ./helm/myapp \
          --namespace pr-${{ github.event.number }} \
          --set image.tag=pr-${{ github.event.number }} \
          --set ingress.host=pr-${{ github.event.number }}.myapp.dev

    - name: Comment PR with env URL
      uses: actions/github-script@v7
      with:
        script: |
          github.rest.issues.createComment({
            issue_number: context.issue.number,
            owner: context.repo.owner,
            repo: context.repo.repo,
            body: `Preview environment: https://pr-${{ github.event.number }}.myapp.dev`
          })

  cleanup:
    runs-on: ubuntu-latest
    if: github.event.action == 'closed'
    steps:
    - name: Destroy ephemeral env
      run: kubectl delete namespace pr-${{ github.event.number }} --ignore-not-found

# Use feature flags instead of environment-specific code paths
from functools import lru_cache
import boto3

@lru_cache(maxsize=1)
def get_flags():
    ssm = boto3.client('ssm')
    return {
        'new_checkout': ssm.get_parameter(Name='/myapp/flags/new_checkout')['Parameter']['Value'] == 'true',
        'ai_recommendations': ssm.get_parameter(Name='/myapp/flags/ai_recommendations')['Parameter']['Value'] == 'true',
    }

# In application code
if get_flags()['new_checkout']:
    return new_checkout_flow(cart)
else:
    return legacy_checkout_flow(cart)

Variable           local           ci              staging         production
DATABASE_URL       sqlite / docker real postgres    RDS instance    RDS (prod)
STRIPE_KEY         test key        test key         test key        live key
OPENAI_API_KEY     dev key         test key         prod key        prod key
DEBUG              true            false            false           false
LOG_LEVEL          DEBUG           INFO             INFO            WARNING

# healthcheck.py — expose /health and /health/dependencies
from fastapi import FastAPI, HTTPException
import asyncpg, redis.asyncio as redis

app = FastAPI()

@app.get("/health")
async def health():
    return {"status": "healthy"}

@app.get("/health/dependencies")
async def dependencies():
    results = {}

    try:
        conn = await asyncpg.connect(DATABASE_URL)
        await conn.execute("SELECT 1")
        await conn.close()
        results["database"] = "ok"
    except Exception as e:
        results["database"] = f"error: {e}"

    try:
        r = await redis.from_url(REDIS_URL)
        await r.ping()
        results["redis"] = "ok"
    except Exception as e:
        results["redis"] = f"error: {e}"

    if any(v != "ok" for v in results.values()):
        raise HTTPException(status_code=503, detail=results)

    return results