Cannot start a span with specified traceId

[dan-cooke]

What happened?

Steps to Reproduce

I have a python service and a node service. they communciate via Redis pub/sub

1. The node service starts a trace and sends it over redis as so

  @Span('DocumentGenerationService.generate')
  async generate(id: number, orgId: string) {
    const span = this.traceService.getSpan()
    const traceId = span.spanContext().traceId;
    console.log("1. Sending TraceID:", traceId);

    this.redis
      .emit('example', { traceId )
      .subscribe({
        error: (err) => console.error(err),
      });
  }

Note: the decorator is from nestjs-otel but it seemingly just starts a span using the default tracer

2. The python service listens on this channel and sends it back

    def handle(self, message: str):
        data = json.loads(message)["data"]
        trace_id_string = data["traceId"]

        print(f"2. Received  traceId: {trace_id_string}")

        span_id = id.generate_span_id()

        span_context = SpanContext(
            int(trace_id_string[:16], 16), span_id=span_id, is_remote=True
        )
        ctx = trace.set_span_in_context(trace.NonRecordingSpan(span_context))

        with tracer.start_as_current_span("python_example", context=ctx):
### --- leaving the rest out as its not important

3. Now my node service listens for this reponse and logs out the trace id

  async example(payload: DocgenResponsePayload) {
    console.log('3. node received traceId: ', payload.traceId);
    const traceIdKey = createContextKey('traceId');
    const doc = await this.documentService.findOne(documentId, orgId);

    const ctx = trace.setSpan(
      context.active().setValue(traceIdKey, payload.traceId),
      this.traceService.getSpan()
    );
    const tracer = trace.getTracer('default');
    const span = tracer.startSpan('onAnswerRulePromptsComplete', {}, ctx);
    console.log('Span context traceId ', span.spanContext().traceId);
}

Now its worth noting i have tired maybe 10 different variations here using all sorts of boilerplate like context.with startActiveSpan etc. etc. and nothing works

Expected Result

All the logs should print the same traceId

Actual Result

1, 2 and 3 print the same traceId => therefore the spans are working on the python side, and I have correctly received the same traceId back

But 4 is printing a new traceId, so when I start a new span on the node service, it will not use the passed traceId in the context

Additional Details

I have been on and off with this for over 2 weeks and I cannot get my node service to work, the documentation for this is extremely confusing and I'm having to dig into ancient github issues to find any examples of this.

I realise most people use a Propagator, but I don't think that will work for redis pub/sub

OpenTelemetry Setup Code

No response

package.json

No response

Relevant log output

No response

[dyladan]

This is the incorrect line context.active().setValue(traceIdKey, payload.traceId),

The trace id alone is insufficient to create a parent context, which requires a trace id and a span id. Interacting with context like this directly using keys is not recommended and is likely to cause issues like what you are experiencing here. Yes, you should be using a propagator for this.