Emit spans which show exponential backoff delay and jitter

[lopopolo-openai]

I have a function that internally uses stamina.retry_context to retry a cross service network call. We use tracing on these requests to allow perf debugging. I've attached below the spans that get emitted by this code. I am fairly certain the gaps between the spans here at least in part are windows where stamina is doing exponential backoff and jitter between the retries, but I am not sure if there is other stuff going on in there too.

Would it be possible to take a callback or something that can be used to get stamina to emit spans for the relevant delays in the retry cycle? It would be excellent if these spans included info about the amount of backoff and jitter, and what the next attempt number will be.

[hynek]

I'm not 100% positive on what you mean by that, but what is missing from https://stamina.hynek.me/en/stable/api.html#stamina.instrumentation.RetryDetails to implement a retry hook?

[hynek]

@lopopolo-openai btw it would be good (for you :)) if we could resolve this fast-ish, because I'm in the last leg of editing a video about stamina and want to release a new version at the same time. :)

[lopopolo-openai]

I'll take a peek on Monday. Usually our tracing code expects spans to be tracked using a context manager.

[hynek]

hmmm, I might be convinced to allow for retry hooks to be context managers but that probably is nothing I can very short term 🤔