data type of metric datapoint value does not match Google Managed Prometheus defined data type

[nabheet]

Component(s)

exporter/googlemanagedprometheus

What happened?

Description

I am getting this error:

One or more TimeSeries could not be written: timeSeries[54]: Value type for metric prometheus.googleapis.com/system_network_errors_total/counter conflicts with the existing value type (INT64).; timeSeries[53]: Value type for metric prometheus.googleapis.com/system_network_errors_total/counter conflicts with the existing value type (INT64).

Steps to Reproduce

I am using Python OTEL instrumentation to generate this data. Here is the detailed debug output from OTEL Collector.

Now, it seems to me that the data coming from the Python Instrumentation contains a float64 value. And I am guessing that in Google Managed Prometheus it is defined as Int64. Which is fine because if figure I could just cast the float to int and drop the decimals as this is a counter and decimals are irrelevant.

So here is what I tried:

  transform/metrics:
    error_mode: ignore
    metric_statements:
      - context: metric
        conditions:
          - resource.attributes["cloud.region"] == nil
        statements:
          - set(resource.attributes["cloud.region"], "${env:GCP_CLOUD_REGION}")

      - context: datapoint
        conditions:
          - metric.name == "system.network.errors"
        statements:
          - set(sum, value_int)
          # - set(value, Int(value))

I don't think I understand OTTL and OTEL configuration enough to implement this. I couldn't find anything specific to what I wanted in the examples. Any assistance would be greatly appreciated.

Expected Result

I would be happy with a way to cast the floating point value to an integer value for each of the system.network.errors metric datapoints.

Actual Result

Please HELP!!!!!

Collector version

v0.112.0

Environment information

Environment

OS: Google Container OS and docker

OpenTelemetry Collector configuration

receivers:
  otlp:
    protocols:
      grpc:
        endpoint: 0.0.0.0:4317

  prometheus/scraper:
    config:
      global:
        scrape_interval: 30s
      scrape_configs:
        - job_name: metric-exporter
          scheme: https
          static_configs:
            - targets:
              - ${env:SOME_METRIC_EXPORTER}

  # prometheus/self-metrics:
  #   config:
  #     scrape_configs:
  #       - job_name: otel-self-metrics
  #         scrape_interval: 1m
  #         static_configs:
  #         - targets:
  #           - localhost:8888

  statsd:
    enable_metric_type: true
    timer_histogram_mapping:
      - statsd_type: "histogram"
        observer_type: "histogram"
        histogram:
          max_size: 50
      - statsd_type: "distribution"
        observer_type: "histogram"
        histogram:
          max_size: 50
      - statsd_type: "timing"
        observer_type: "summary"

processors:
  batch:

  memory_limiter:
    # drop metrics if memory usage gets too high
    check_interval: 10s
    limit_percentage: 95
    spike_limit_percentage: 20

  # automatically detect Cloud Run resource metadata
  resourcedetection:
    detectors:
      - env
      - gcp

  filter/self-metrics:
    error_mode: ignore
    metrics:
      include:
        match_type: strict
        metric_names:
        - otelcol_process_uptime
        - otelcol_process_memory_rss
        - otelcol_grpc_io_client_completed_rpcs
        - otelcol_googlecloudmonitoring_point_count

  filter/drop-metrics:
    error_mode: ignore
    metrics:
      metric:
        - name == "http.server.response.size"
        - name == "http.server.request.size"
        - name == "http.client.response.size"
        - name == "http.client.request.size"


        - name == "process.open_file_descriptor.count"
        - name == "process.runtime.cpython.cpu.utilization"
        - name == "process.runtime.cpython.thread_count"
        - name == "process.runtime.cpython.memory"
        - name == "http.client.duration"

  transform/metrics:
    error_mode: ignore
    metric_statements:
      - context: metric
        conditions:
          - resource.attributes["cloud.region"] == nil
        statements:
          - set(resource.attributes["cloud.region"], "${env:GCP_CLOUD_REGION}")

      - context: datapoint
        conditions:
          - metric.name == "system.network.errors"
        statements:
          - set(sum, value_int)

exporters:
  googlemanagedprometheus: # Note: this is intentionally left blank
  googlecloud:             # Note: this is intentionally left blank
  debug:
    verbosity: detailed

extensions:
  health_check:
    endpoint: "0.0.0.0:${env:OTEL_HEALTHCHECK_PORT}"

service:
  extensions:
    - health_check
  telemetry:
    logs:
      level: debug
  pipelines:
    traces:
      receivers:
        - otlp
      processors:
        - memory_limiter
        - batch
      exporters:
        - googlecloud

    metrics:
      receivers:
        - otlp
        - prometheus/scraper
        - statsd
      processors:
        - memory_limiter
        - filter/drop-metrics
        - transform/metrics
        - batch
      exporters:
        - googlemanagedprometheus
        # leave this debug exporter commented out so it is
        # easier to enable and I don't forget about it
        - debug

    # metrics/self-metrics:
    #   exporters:
    #     - googlemanagedprometheus
    #   processors:
    #     - filter/self-metrics
    #     - memory_limiter
    #     - resourcedetection
    #     - batch
    #   receivers:
    #     - prometheus/self-metrics

Log output

No response

Additional context

No response

[github-actions]

Pinging code owners:

• exporter/googlemanagedprometheus: @aabmass @dashpole @jsuereth @punya @psx95

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[dashpole]

Take a look at https://github.com/open-telemetry/opentelemetry-collector-contrib/tree/main/exporter/googlemanagedprometheusexporter#conflicting-value-types. That should help you solve it. Let me know if you have any other questions

[nabheet]

Perfect! thank you! This seems to have solved this particular error.