Add Testing for Shutdown Scenarios. about opentelemetry-rust HOT 2 OPEN

Looking into this, I'd love some clarification on how the scenario would work. My current approach has been to set up an integration test based on https://github.com/open-telemetry/opentelemetry-rust/blob/main/examples/logs-basic/src/main.rs and extend it such that we have multiple background tasks with set up with references to the LoggerProvider instance through an Arc.

The shutdown I wanted to test was to create a forced shutdown through an interrupt. I initially tried this with setting up multiple exporters with a single LoggerProvider and got a panic from this line

Would taking the same approach with multiple LoggerProvider references in separate tasks be a similar scenario?

from opentelemetry-rust.

The shutdown I wanted to test was to create a forced shutdown through an interrupt. I initially tried this with setting up multiple exporters with a single LoggerProvider and got a panic from this line

opentelemetry-rust/opentelemetry-sdk/src/logs/log_emitter.rs

Line 86 in 36c2a8e

.expect("cannot shutdown LoggerProvider when child Loggers are still active")

which I think is expected behavior.

Would taking the same approach with multiple LoggerProvider references in separate tasks be a similar scenario?

Looking at our API and the spec I actually think we might need a change. https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/logs/sdk.md#shutdown

Right now i think this panic isn't actually doing what I would want/expect. There's already a result being returned. In this case it's a vec of results but I feel like we should be passing back a default of Vec error result.

Shutdown SHOULD provide a way to let the caller know whether it succeeded, failed or timed out.

Panics don't really "let the caller know".

from opentelemetry-rust.