OpenTelemetry

The exporter is OpenTelemetry-first: in addition to the Prometheus /metrics endpoint it can push metrics via OTLP, and optionally emit traces for diagnosing slow collection cycles.

Metric push

opentelemetry:
  metrics:
    enabled: true
    endpoint: "otel-collector:4317"
    insecure: true
    interval: "10s"

When enabled, the exporter registers an observable gauge per metric name; a periodic reader collects them every interval and pushes to the OTLP gRPC endpoint. The same in-memory snapshot feeds both the OTLP push and the Prometheus scrape, so the two paths always agree.

Every data point carries the same attributes as the Prometheus labels, including cluster. Metric names match the Prometheus names (e.g. pflex_cluster_iops).

Tracing

opentelemetry:
  tracing:
    enabled: true
    endpoint: "otel-collector:4317"
    insecure: true
    samplingRate: 0.1

Tracing instruments each collection cycle (collect.cycle) and the per-request PowerFlex API calls (powerflex.request), which helps pinpoint a slow gateway or a cluster that is dragging out a cycle. Tracing is independent of the metric push — enable either, both, or neither.

Collector example

otel-collector-config.yaml (used by the Compose stack) accepts the OTLP push and re-exposes the metrics for Prometheus:

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