[Monitor Opentelemetry] Question about ensuring trace correlation in Azure Functions

[gunzip]

I'm trying to achieve proper trace correlation in an Azure Function using the NodeJS worker.

The Azure Function runtime (host) logs the incoming HTTP request, while the NodeJS worker uses the Application Insights SDK (v3.x), which in turn leverages useAzureMonitor configured with the ApplicationInsightsSampler.

It appears that this sampler may drop traces related to the Span Context opened by the Azure host (root HTTP request), as it does not retain samples based on parent spans. This could result in incomplete end-to-end traces when using Azure Functions even with the same sampleRatio configured for both host and worker.

Would it be possible to use a ParentBasedSampler instead, to retain the full chain when the host sample the original request?

[gunzip]

Update: I’ve reviewed the sampling algorithm used in donet (Azure Function Host), and it invokes the DJB2 hashing function as follows:

https://github.com/Azure/azure-sdk-for-net/blob/main/sdk/monitor/Azure.Monitor.OpenTelemetry.Exporter/src/Internals/ApplicationInsightsSampler.cs#L57

while in the JS implementation it looks like it's applied directly to the plain (lowercase) traceId value:

https://github.com/Azure/azure-sdk-for-js/blob/main/sdk/monitor/monitor-opentelemetry/src/traces/sampler.ts#L71

Could this be the reason?