I got OpenTelemetry to work. But why was it so complicated?

Re:I got OpenTelemetry to work. But why was it so complicated? by csomar on Saturday January 11, 2025 @07:25AM (#42665422)

Otel is a spec. You can create your own clients/aggregators/etc.. The problem is that if nobody does it, there will be no tooling. So Otel created some tooling (and yes, it's bad) for people to use.

Some companies (ie: Datadog) are contributing to the tooling but I think most companies would rather spend dev time on their own platforms than something that anybody (competitor) can use.

[ Parent | on HN ]

Re:I got OpenTelemetry to work. But why was it so complicated? by bluesnews on Saturday January 11, 2025 @12:20PM (#42667312)

From the user side, a spec isn't helpful unless it has implementations. And the official implementations are complicated compared to prometheus.

[ Parent | on HN ]

Re:I got OpenTelemetry to work. But why was it so complicated? by MathMonkeyMan on Saturday January 11, 2025 @03:14PM (#42668606)

I worked on a team that produced a distributed tracing library. We were tasked with interoperating with OpenTelemetry, or at least figuring out what that means.

My teammate said that at a previous job he wanted to add OpenTelemetry tracing to some C++ code he was working on. He took one look at the reference implementation for C++ OpenTelemetry and decided instead to write his own tracing library that sends gRPC to the OpenTelemetry collector.

It's also worth noting that, at least last time I checked, the reference implementations per programming language are less like reference implementations of some specification, and more like "this is the code you use to do OpenTelemetry in this language."

[ Parent | on HN ]