railway-deployments

Railway Deployments (awfixersites)

Deploy methods

Method	When
Terraform	Durable service config (Postgres, workers)
CLI railway up	Ad-hoc deploy from linked directory
MCP deploy	Agent-driven deploy from natural language
MCP redeploy (remote)	Restart without code change
Git push	If service has Railway GitHub integration

CLI deploy

railway link          # ensure correct project/service
railway up            # deploy current directory
railway up --detach   # don't stream logs

Verify deployment

railway status
railway logs
railway logs --follow
railway open          # dashboard

Failure diagnostic tree

Deploy failed?
├─ Build error
│  ├─ Check Dockerfile / nixpacks config
│  ├─ Missing env var? → railway variables
│  └─ Wrong root directory? → check railway.toml
├─ Runtime crash
│  ├─ railway logs --follow
│  ├─ DATABASE_URL missing? → check Postgres service + variables
│  └─ Port binding? → Railway expects $PORT env var
├─ Health check failure
│  ├─ Service not listening on $PORT
│  └─ Startup too slow → increase health check timeout
└─ Volume / data issue
   ├─ Postgres not ready → wait for template init
   └─ Connection refused → check internal networking

For complex failures, use remote MCP railway-agent tool.

awfixersites context

• Next.js apps deploy to Vercel, not Railway
• Railway deploys: Postgres, future workers, webhook receivers
• Never deploy a Next.js app to Railway without explicit user request
• Production deploys require user confirmation

Redeploy (no code change)

# Via remote MCP
redeploy <service> in production

# Via dashboard
Project → Service → Deployments → Redeploy