Counter metric resets on every refresh

[manu-balakrishnan]

Hi,

  I am implementing opentelemetry counter metric on react JS application. it seems metric values getting reset to zero (or re-initializing) on every refresh and start updating its value from zero instead of previous value. Is there any way to maintain the counter value without resetting to zero?

[legendecas]

Would you mind sharing your setup code for OpenTelemetry SDK? My first thought on this is that you are using a delta exporter, but rather you are expecting the exporter to use the cumulative aggregation temporality.

[manu-balakrishnan]

@legendecas Please let me know if something I am missing to configure. Your support will be much appreciated.

[legendecas]

@manu-balakrishnan I noticed you are using a collector to transport the metric data to the Prometheus backend. By default, the JS collector exporter collects cumulative data points (with AggregationTemporality.CULUMATIVE), and I tested the setup code you shared with a local OTLP collector logging output and verified that the collector is outputting the expected cumulative metrics data points.

Would you mind sharing your OLTP collector version and configurations on how the collector sends the metrics to your Prometheus backends? Also, if possible, you can configure a logging output to verify the metrics data the collector receives: https://github.com/open-telemetry/opentelemetry-collector/blob/main/exporter/loggingexporter/README.md.

[manu-balakrishnan]

Hi @legendecas, Please find the docker compose yaml and collector configuration yaml for your reference.

docker-compose.yml :

version: "3"
services:
  # Collector
  collector:
    image: otel/opentelemetry-collector-contrib:0.62.0
    command: ["--config=/conf/collector-config.yaml"]
    volumes:
      - ./collector-config.yaml:/conf/collector-config.yaml
      - ./output:/etc/output:rw
    ports:
      - "9464:9464"
      - "55681:55681"
      - "4318:4318"
      - "4317:4317"
      - "8888:8888"   # Prometheus metrics exposed by the collector
      - "8889:8889"   # Prometheus exporter metrics
    depends_on:
      - jaeger-all-in-one

  # Jaeger
  jaeger-all-in-one:
    image: jaegertracing/all-in-one:latest
    ports:
      - "16686:16686"
      - "14268"
      - "14250"
  prometheus:
    container_name: prometheus
    image: prom/prometheus:latest
    volumes:
      - ./prometheus.yaml:/etc/prometheus/prometheus.yml
    ports:
      - "9090:9090"

collector-config.yml:

receivers:
  otlp:
    protocols:
      http:
        endpoint: 0.0.0.0:4318
        #cors_allowed_origins: 
        #  - http://localhost:3000
        cors:
          allowed_origins:
            - http://*
        #    - https://*

exporters:
  jaeger:
    endpoint: jaeger-all-in-one:14250
    #insecure: true
    tls:
      insecure: true
  prometheus:
    endpoint: "0.0.0.0:9464"
  file:
    path: /etc/output/logs.json
processors:
  batch:

service:
  pipelines:
    traces:
      receivers: [otlp]
      exporters: [jaeger, file]
      processors: [batch]
    metrics:
      receivers: [otlp]
      exporters: [prometheus]
      processors: [batch]

As I mentioned above, here the issue is whenever we refresh the page, metric value getting reset to zero and start updating from zero to one. until and unless page refreshes, counter metric updating its value as expected.

[manu-balakrishnan]

logs.zip
@legendecas Attaching the logs of the same for reference.

[legendecas]

Sorry about missing the keyword "refresh the page". This behavior is by design -- OpenTelemetry library doesn't save any metric states after the page refreshes.

Backends like Prometheus are able to detect metric resets and they can calculate the correct result of that metric, e.g. https://prometheus.io/docs/prometheus/latest/querying/functions/#resets.