Populate message key in kafka-exporter

[findingrish]

Currently the paltform receives messages with key as null, can the collector set the key as TenantId (default value in case it is not set) + TraceId instead (similar to https://github.com/hypertrace/hypertrace-ingester/blob/main/span-normalizer/span-normalizer/src/main/java/org/hypertrace/core/spannormalizer/jaeger/JaegerSpanToAvroRawSpanTransformer.java#L50).

1. There is going to be no downsides of doing this
2. This could potentially allow us to directly group the span into Traces in the entry platform service (Remove span-normalizer service  hypertrace-ingester#59).

cc @jcchavezs @buchi-busireddy @kotharironak

[pavolloffay]

Currently the paltform receives messages with key as null,

by currently do you mean when using the OpenCensus service or this collector?

[findingrish]

by currently do you mean when using the OpenCensus service or this collector?

OC collector

[pavolloffay]

Alright so this is a new feature request. I thought there is a regression between OC kafka exporter and OTEL collector.

[pavolloffay]

Would it be enough to use the TraceID without tenant as the key?

The kafka exporter code is in OTEL collector. Adding traceID as a key might seem feasible, however the collector upstream does not have a concept of tenantID.

[jcchavezs]

yeah, tenant_id feels more like a non hypertrace concern.