Railway Deployment

Railway Deployment

Guide Railway deployments from preflight validation to post-deploy verification.

When to Use

• You are preparing to deploy or update an application on Railway.
• You need a repeatable deployment checklist with explicit verification.
• You want to minimize release risk with rollback planning.

Workflow

1. Confirm deployment target and intended version/change scope. Identify the project and service(s). Choose deployment region (e.g., US West, EU) if relevant. Note whether this is a new deploy or an update.

2. Validate runtime and configuration. Use railway.json or Procfile to define build and start commands explicitly; do not rely on Nixpacks guessing. Example railway.json: {"build": {"builder": "NIXPACKS"}, "deploy": {"startCommand": "npm run start", "restartPolicyType": "ON_FAILURE", "restartPolicyMaxRetries": 3}}. For databases, ensure TCP proxy is enabled if connecting from external clients; Railway exposes DATABASE_URL with internal hostnames.

3. Configure volume mounts and environment groups. If the app needs persistent storage, add a volume and mount path in the Railway dashboard. Use environment variable groups to separate staging from production. Reference variables with $VAR syntax in railway.json or dashboard.

4. Identify risky changes. Review schema migrations, breaking env changes, and external API dependencies. Plan migration order (e.g., run migrations before deploy or use backward-compatible migrations).

5. Define verification plan. Ensure a health check endpoint exists and returns 200. Railway does not require a health check path for basic deploys, but configuring one improves reliability. Plan checks for critical user flows and error rate.

6. Execute deploy and monitor. Use railway up or connect a GitHub repo for automatic deploys. Watch build and deploy logs in the dashboard or via railway logs. On Starter plan, services may sleep after inactivity; document this behavior.

7. Trigger rollback if checks fail. Use Railway's deployment history: Dashboard -> Service -> Deployments -> select previous deploy -> Redeploy. Capture incident notes for post-mortem.

Common Pitfalls

• Not setting start command. Nixpacks may guess incorrectly (e.g., wrong entry point or missing flags). Always set startCommand in railway.json, Procfile, or dashboard.
• Missing health check endpoint. Without a health endpoint, Railway cannot detect unhealthy instances; consider adding /health or similar for monitoring and load balancer checks.
• Ignoring Railway sleep on Starter plan. Services sleep after inactivity; first requests after sleep can be slow. Use a paid plan or external cron to keep services warm for production.
• Database connection pooling with serverless. If using serverless runtimes (e.g., serverless functions), connection pooling can exhaust database connections. Use connection poolers (e.g., PgBouncer) or serverless-friendly drivers.

Railway CLI and Configuration

railway login
railway link          # Link to existing project
railway up            # Deploy from current directory
railway logs          # Stream logs
railway run <cmd>     # Run command with env vars injected
railway variables     # List env vars

Example railway.json:

{
  "$schema": "https://railway.app/railway.schema.json",
  "build": {
    "builder": "NIXPACKS",
    "buildCommand": "npm run build"
  },
  "deploy": {
    "startCommand": "npm run start",
    "healthcheckPath": "/health",
    "healthcheckTimeout": 30,
    "restartPolicyType": "ON_FAILURE",
    "restartPolicyMaxRetries": 3
  }
}

Example Procfile: web: npm run start

Rollback and Monitoring

Railway stores deployment history per service. Rollback: Dashboard -> select service -> Deployments -> choose previous successful deploy -> Redeploy. Use railway logs for real-time debugging. Integrate with external monitoring (e.g., Sentry, Datadog) for error tracking and performance.

Output Format

## Deployment Plan
- Target service: <name>
- Change scope: <summary>
- Region: <if applicable>
- Start command: <explicit command>

## Preflight Checklist
- [ ] Required env vars validated
- [ ] Runtime/build settings confirmed (railway.json or Procfile)
- [ ] Migration risk reviewed
- [ ] Health check endpoint configured
- [ ] Volume mounts configured if needed

## Commands
1. railway up
2. railway logs

## Verification
- [ ] Health endpoint passes
- [ ] Key user flow validated
- [ ] Error rate and logs normal
- [ ] Sleep behavior documented (Starter plan)

## Rollback
- Trigger: <condition>
- Action: Dashboard -> Deployments -> select prior deploy -> Redeploy
- Notes: <incident capture>

Constraints

• Prefer deterministic checklists and explicit pass/fail criteria.
• Do not assume deploy success until verification steps are complete.
• Highlight unknowns instead of guessing environment details.
• Always set startCommand explicitly; do not rely on Nixpacks detection alone.
• Document sleep behavior for Starter plan services.