OpenTelemetry

Besides the Prometheus /metrics pull endpoint, the exporter can push metrics (and optionally traces) over OTLP/gRPC. Both export paths read the same shared snapshot, so the push cadence is independent of the collection interval and of any Prometheus scrapers.

Configuration

opentelemetry:
  metrics:
    enabled: true
    endpoint: "otel-collector:4317"
    insecure: true
    interval: "30s"
  tracing:
    enabled: false
    endpoint: "otel-collector:4317"
    insecure: true
    samplingRate: 0.1

Field	Meaning
metrics.enabled	Turn on the OTLP metric push.
metrics.endpoint	OTLP/gRPC collector address (host:port).
metrics.insecure	true for plaintext gRPC (lab/in-cluster); use TLS otherwise.
metrics.interval	How often metrics are pushed.
tracing.enabled	Emit OTLP traces for the collection loop.
tracing.samplingRate	Fraction of traces sampled (0.0–1.0).

The metric push uses asynchronous observable gauges driven by a periodic reader that reads the latest snapshot — the same data served on /metrics, so series names and labels match the Metrics Reference.

Collector pipeline

The bundled otel-collector-config.yaml receives OTLP on :4317 and re-exposes the metrics on :8889 for Prometheus to scrape:

receivers:
  otlp:
    protocols:
      grpc:
        endpoint: 0.0.0.0:4317
processors:
  batch: {}
exporters:
  debug:
    verbosity: normal
  prometheus:
    endpoint: 0.0.0.0:8889
service:
  pipelines:
    metrics:
      receivers: [otlp]
      processors: [batch]
      exporters: [debug, prometheus]

End-to-end in the compose stack

The local stack wires this up: enable opentelemetry.metrics pointing at otel-collector:4317, recreate the exporter, and Prometheus picks up the re-exposed series via its powerscale-otlp job (otel-collector:8889). This lets you observe both export paths side by side — powerscale (direct pull) and powerscale-otlp (push → collector → scrape).