This ticket was split from DRIVERS-719, please see that ticket for a detailed description.

Questions and answers posed by the spike in PYTHON-5744:

• Determine if the existing opentelemetry-instrumentation-pymongo package can be used as the underlying implementation for PyMongo's OpenTelemetry support in accordance with the spec.
  • Answer: it cannot. The spec requires different functionality than the package supports. Additionally, relying on the existing package would not allow useful consolidation of our telemetry APIs.
• If opentelemetry-instrumentation-pymongo isn't suitable for implementing the spec, can we contribute to it upstream to make it suitable? Or are we better off creating our own internal solution?
  • Answer: We are better off creating our own internal solution.
• The spec details specific OpenTelemetry span attributes to be recorded for each driver operation. What does that recording look like over the lifecycle of a PyMongo operation?
  • Answer: Spans are begun when an operation begins, and are completed when that operation finishes either successfully or with an error. Our proposed unified telemetry API would make this tracking simple.
• Can we use our existing debug logging code to guide our OpenTelemetry support?
  • Answer: Yes. We plan to unify our telemetry APIs so that a single call is used to record for any of the enabled telemetries.
• Does abstracting both debug logging and OpenTelemetry logging into a single, centralized design make sense? Both appear to gather similar information at similar points, reducing duplication would be ideal.
  • Answer: Yes.