OTEL was designed like it’s a Java app, and I mean that as a condemnation.
The errors are opaque and the API is obtuse. It’s also not quite done yet. The Node implementation has had major major bugs only fixed in the last twelve months, and necessary parts of the ecosystem (across all languages) are still experimental.
Overengineered and underdelivering. I’m using Prometheus on my next project. Though I’m tempted to go back to StatsD if I can find a good tagged version. Don’t hold data in a process with an unknown lifetime. Send it over loopback to a sidecar immediately and let it sort aggregation out.
With this approach you’re standing on the shoulders of giants, compatible with any agent that speaks OTLP, and can market your distribution as an ecosystem tool.
This is clearly the industry standard protocol and the present and future of o11y.
The whole point is that o11y vendors can stop reinventing lower level protocols and actually offer unique value props to their customers.
So why would you want to waste your time on such an endeavor?