opentelemetry/exporter-metrics-otlp-http Does not send metrics to collector (with last collector version 0.40.0)

[Nighttraveler]

Hello guys, We wanted to use https://github.com/GoogleChrome/web-vitals#browser-support to get some metrics from our browser instrumentation and send them through opentelemetry-js. In order to accomplish that we started to make some tests with the examples provided here.

What version of OpenTelemetry are you using?

I'm using 0.27.0 opentelemetry-js version (this repo in main branch) and otelcontribcol 0.40.0

What version of Node are you using?

v12.16.3

Please provide the code you used to setup the OpenTelemetry SDK

At the moment the exporter does not work with the examples provided.
Just changed this in the example (tracer-web/metrics/index.js)
const labels = { pid: 1, environment: 'staging' }; because 'process' is undefined

What did you do?

We just simply start up the 'tracer-web' examples.
All the traces examples work as expected and arrive to our collector, but metrics doesn't.

If possible, provide a recipe for reproducing the error.

• Have your collector running
• cd opentelemetry-js/examples/tracer-web
• npm install
• in metrics/index.js change:
  const labels = { pid: process.pid, environment: 'staging' }; to const labels = { pid: 1, environment: 'staging' };
• npm run start
• Go to http://localhost:8090/metrics/ and click 'Start metrics'

What did you expect to see?

I expected to see the metrics in our collector/backed, like all the other traces examples that works correctly.

What did you see instead?

In Network tab from the browser I have a 400 Bad request response from the collector.

{"code":3,"message":"unknown field \"doubleSum\" in v1.Metric"}

Additional context

• traces are exported to localhost:55681/v1/traces (default legacy endpoint)

• metrics are exported to localhost:55681/v1/metrics (default legacy endpoint)

• Docker compose collector file:

version: "2"
services:

  # Collector
  otel-collector:
    image: otel/opentelemetry-collector-contrib:0.40.0
    command: ["--config=/etc/otel-collector-local-config.yaml"]
    volumes:
      - ./otel-collector-local-config.yaml:/etc/otel-collector-local-config.yaml
      - /home/fernando/workspace/khoros/agent/collector-volume:/etc/collector-logging/
    environment:
      LOG_EXPORTER_LOG_LEVEL: "DEBUG"
      NEW_RELIC_API_KEY: "key123"
    ports:
      - "1888:1888"   # pprof extension
      - "13133:13133" # health_check extension
      - "4317:4317"   # OTLP gRPC receiver
      - "4318:4318"   # OTLP HTTP receiver
      - "55681:55681" # default legacy OTLP HTTP receiver
      - "55670:55679" # zpages extension
      - "8888:8888"   # Prometheus metrics exposed by the collector
      - "8889:8889"   # Prometheus exporter metrics

• Collector yml config file:

receivers:
  otlp:
    protocols:
      grpc:
      http:
        cors_allowed_origins:
        - http://*
        - https://*        
exporters:
  logging:
    logLevel: $LOG_EXPORTER_LOG_LEVEL
  otlp:
    endpoint: https://otlp.nr-data.net:4317
    headers:
      "api-key": $NEW_RELIC_API_KEY
processors:
  batch:
extensions:
  health_check: {}
service:
  extensions: [health_check]
  pipelines:
    traces:
      receivers: [otlp]
      processors: [batch]
      exporters: [otlp, logging]
    metrics:
      receivers: [otlp]
      exporters: [otlp, logging]

Hope the information is enough to reproduce the issue, if not, please reach me and I provide more details.

[morigs]

Same issue here. Also tested with otelcol 0.41.0, otelcontribcol 0.40.0 and otelcontribcol 0.41.0

[morigs]

I believe this is due to inconsistencies between payload of the Metrics and actual proto type.

This is caused by this change. This changed in proto 0.9.0 which delivered to the collector in version 0.33.0.

Perhaps fix is quite easy: just rename fields to the new scheme and remove unused (int) fields.

[dyladan]

Yeah our proto is a little out of date. i'll update it

[morigs]

Looking forward! If any help is needed, just let me know

[dyladan]

I won't have time to tackle this until monday so if you want to do it before then I will try to keep an eye out for notifications and merge/release the fix over the weekend. If you can wait until monday i'll just handle it then.

[morigs]

@dyladan sorry for pinging you, is there any success?

[RomeroGaliza]

Same issue here!

[morigs]

@dyladan sorry for disturbing you again. Could you pls add help-needed badge or something. Just to draw attention of possible contributors to the issue.
Or maybe you can provide a high-level list of steps required for proto upgrade, so I can do it myself? That would be awesome

[dyladan]

In order to get this working, the proto transformations need to be updated. The transformations are in experimental/packages/opentelemetry-exporter-metrics-otlp-http/src/transformMetrics.ts and depend on the types in packages/exporter-trace-otlp-http/src/types.ts. Because of this unfortunate coupling, the proto files, their types (hand rolled), and the transformations (also hand rolled) need to be updated for all 6 otlp exporters at the same time. If you want to get the fix out ASAP, I would recommend you do that.

Alternatively, I am working on a separate package which will handle transformations into the intermediate OTLP JSON format and serialization as protobuf in #2691. Once that package is completed, the metrics exporters will be refactored to use it (and it already has updated proto). Depending on how time-sensitive this is for you, it may be prudent to just wait for that (i would estimate a couple weeks).

[mbecca]

Hello, is there any progress?

[sgracias1]

I ran a test with the OTLP trace exporter and noticed it fails to connect to the collector for version > 0.42.0. Does this issue track the trace exporter as well, or should that be a separate issue?