Feat(180): backend metrics support

[brainrepo]

closes #180

• [x] Update docker compose for spin up prometheus
• [x] Update collector to expose the metrics for prometheus
• [x] Create setup function for configuring The Open Telemetry metrics
• [x] Create a plugin that allow to expose these metrics
  • Counter for request / path
  • Gauge of the current active requests
  • Histogram for the response times

[claudia-codes]

How do you tests this?

[brainrepo]

@claudia-codes

• start containers with docker-compose up
• visit http://127.0.0.1:3000/health a bunch of times
• http://127.0.0.1:9464/metrics shuld show your http_server_active_requests, http_server_requests and http_server_request_times_bucket
• Prometheus reachable at http://127.0.0.1:9090/ should allow to visualize the metrics

[brainrepo]

hey @brainrepo it seems after the tracing refactor it doesn't work as expected. When you go to jaeger (once docker compose is up) go to http://0.0.0.0:16686/ you should see traces from front-end app correlated with back-end requests, but it's not the case.

I didn't dive into the refactor logic, but if you'll be stuck we can pair on that if needed :)

Hi @zimny, I found a solution. To share the context correctly, registerInstrumentations have to be called from the global scope and not from a plugin, doing that the traces are correlated, but I can't find the reason why, in the docs, this edge case is not mentioned, would you happen to have any clue?
Ps. All the examples available run registerInstrumentations from the global scope.

[zimny]

Yes, it's normal as open telemetry is doing some sort of monkey patching and it has to be started "first". This workaround seems ok for me. Good to go!