Distributed tracing is hard! That is a statement that most folks would agree too. At large organizations like eBay where there are 1000s of microservices and call chains that have 100s or even 1000s of spans, the complexity to adopt tracing is fairly high. Over the course of the last couple of years, we have traveled far in getting to a state where most applications are instrumented for tracing. However, the journey has numerous hurdles that we have had to cross. This talk discusses, a solid approach we took with the help of Open Telemetry, how it jump started our adoption but at the same time, how we had to do more to get much closer to the finish line.
Key points that we would discuss include:
- Our strategy to deliver instrumentation to all applications
- Challenges in context propagation
- Doing more than what open telemetry alone provides
- Experiences required to make tracing useful
- Focus for the future
