It has been a while since I’ve been working as SRE/Platform/Cloud Engineer, and lately and I realize I’ve been repeating some questions to developers that I rarely get an answer for straight away.

These are not meant to make anyone’s life harder, au contraire, the whole pourpose of having a solid answer to this list of questions, is to make everyone less worried about the probabilty of some high stakes, overnight failure or a data handling missuse that could potentially cause big losses, and of course a lot of unnecessary stress.

If you’re a Software Engineer/Developer, then consider that a service (at least, for me), is a piece of code running in a live production system, that YOU wrote, only YOU know how it works, thus YOU own.


List of questions I’d like to get an answer from devs on any new service deployment request:

  • How can I check the health of the service ?

    • Did you securely configured endpoints that I can ping periodically?
    • Is it going to use the same port as the service ?
  • How can I safely and gracefully restart the service ?

    • Does graceful shutdowns wait for inflight requests to finish?
    • If I restart the service, will there be any disruption or performance degradation?
  • How and why would the service fail ?

    • Does it has any external dependencies ?
    • And what happens if it does ?
    • Do you have a playbook, or sequence of steps, to bring the service back up?
  • Do you use appropriate logging levels depending on the environments ? i.e. (TRACE & DEBUG should not go to production)

    • Are you logging to stdout ?
    • What format of logs are you using, (JSON, PLAINTEXT)?
  • What kind of metrics are you exposing ?

    • Are you measuring the RED signals?
      • Rate (# of requests per second)
      • Errors (# of errors per second) and
      • Duration (time each request takes)
  • Is there any documentation/design specification for the service ?

    • Are you using gRPC or REST ?
    • How can I see the API contract?
    • Do you have a OpenAPI/Swagger spec
  • How does the data flow through the service ?

    • Do you have any PII/Sensitive data flowing through the service?
    • Is any part of the data captured in the service logs?
  • What is the testing coverage for this service ?

    • Do you wrote unit, integration tests for it?
    • Is there a end to end test that I can run discard issues?

My conundrum lies on the following:

  • Luckily, much of this overhead can be alleviated with a service template that I can control and upgrade overtime.

  • But… that same template can abstract and hide away these answers from devs, causing them to ignore issues and fail to react when their services fail.