ReleaseEngineering/Development Best Practices

Language

• Python
• JS

Framework

• web:
  • flask
• db api:
  • sqlalchemy
• daemons:
  • don't manage, use supervisord
  • not async if you can help it
    • shell out to celery
  • tasks: ?
  • general:
    • requests

Resiliency

• retry everything, fail gently
• no in-memory state (including call stack)
• support multiple endpoints (db, redis, rabbitmq, etc) [e.g. no single-point-of-failure]
• sync + caching, you have slightly stale data but still keep running if the service you need to query is down
• run multiple copies of the app at once without conflicts

Deployment

• db:
  • clustered mysql - innodb vs myisam.
  • nosql - redis
• mq:
  • rabbitmq
• logging:
  • log to stdout (use python logging)
• use supervisord
• deploy python packages
• semantic versioning (semver.org)

Security

• passwords
• secrets

Performance

• multi-threading
• scalability
• configuration