Go Production Checklist

This is a work-in-progress checklist for go-services to keep in mind before going to production. If you feel like something is missing, please open an issue or PR.

Startup and shutdown

• Exit early (panic) on fatal errors (e.g. database not found or configuration invalid)
  • Check that all configuration values are present and valid on startup
  • Ping required services (DB, 3rd party APIs, ...) on startup to verify correct configuration and valid credentials
• Migrate your database on startup
• Implement graceful-shutdown unless your gateway does that for you

Code

• Use Dep to manage your dependencies (and install them on bu
• Configure your logger to output structured machine-readable logs, to be processed later (e.g. with ELK)
• Configure a timeout for all http.Clients (the default client has none)
• Close all ReadClosers (e.g. http.Client#Do returns a body that has to be closed manually)
• Implement health-check
• Add CORS if your services exposes an API

Middleware

• Implement circuit breaker if needed
• Implement rate-limiting if needed
• Set up sensible metrics (error-rate, response-time, etc.)
• Enforce body-size-limits if your gateway is not already doing this

Preparation

• Load-test and/or benchmark the service, especially memory footprint
• Verify that your production configuration is valid (this should be the only thing different to staging, so make sure it’s correct)

Deployment

• Use a minimal Dockerfile (multi-stage) if you use Docker
• Compile your service(s) w/ -buildmode=pie (position independent executables)
• Configure SSL so that you get an A+ on SSL Labs if you use SSL