forceFlush does not flush synchronously, should it?

[ktoso]

Need to check what otel expects here but it tripped up some developers using swift-otel in CLI like situations so they didn't get their spans.

Should we also flush during a graceful shutdown, integrated with swift-service-lifecycle?

[slashmo]

Thanks for bringing this up!

Quoting the OTel spec:

This method provides a way for provider to immediately export all spans that have not yet been exported for all the internal processors.
ForceFlush SHOULD provide a way to let the caller know whether it succeeded, failed or timed out.
ForceFlush SHOULD complete or abort within some timeout. ForceFlush can be implemented as a blocking API or an asynchronous API which notifies the caller via a callback or an event. OpenTelemetry client authors can decide if they want to make the flush timeout configurable.
ForceFlush MUST invoke ForceFlush on all registered SpanProcessors.

https://github.com/open-telemetry/opentelemetry-specification/blob/v1.20.0/specification/trace/sdk.md#forceflush

I guess we have two options here. Either we change the Tracer API to make it async or make the implementation blocking in Swift OTel. The first option is not very feasible I guess because it would break API. The second option feels weird when the rest of the API is async, but possible.

---

Should we also flush during a graceful shutdown, integrated with swift-service-lifecycle?

AFAIK it's up to the individual processor how to handle shut down. In the batch span processor e.g. I do force-flush before shutting down the exporter.

swift-otel/Sources/OTel/Tracing/Processing/Batch/OTelBatchSpanProcessor.swift

Line 68 in 83b25c2

try? await forceFlush()