imandra-ai / ocaml-opentelemetry

Instrumentation for https://opentelemetry.io
http://docs.imandra.ai/ocaml-opentelemetry/
33 stars 7 forks source link

simplify sync ocurl backend #33

Closed c-cube closed 1 year ago

c-cube commented 1 year ago

this needs testing, and it removes options, but it's a lot more simple conceptually and in the code.

A single thread collects, via a queue, signals coming from the instrumented application. It then batches signals according to their type, alongside a batch-level timestamp, and sends the whole batch at once if it became big enough or old enough. Sending is performed on a pool of background threads that use ezcurl.

c-cube commented 1 year ago

It's currently deployed on imandra-perf-watch, which seems to still be emitting on datadog. I also tested locally a bit.