Currently when the megaexample calls itself via HTTP it doesn't pass the span context in headers like it used to in pre-eventlog days. Nor does it try to extract the span context when receiving an HTTP request. This stops us from seeing all the spans belonging to a single trace.
The root issue here is that since the eventlog migration the instrumented application doesn't even have any trace ids or span ids in memory, those are generated later in the exporter process.
ethercrow
commented
4 years ago
Currently when the megaexample calls itself via HTTP it doesn't pass the span context in headers like it used to in pre-eventlog days. Nor does it try to extract the span context when receiving an HTTP request. This stops us from seeing all the spans belonging to a single trace.
The root issue here is that since the eventlog migration the instrumented application doesn't even have any trace ids or span ids in memory, those are generated later in the exporter process.