Runtime telemetry (agent)Copy Anchor Link

Cluster workloads instrumented through the Helm chart can feed Observability → Runtimes without embedding Owlpane SDKs in every image.

Data pathCopy Anchor Link

Pod (OTel auto-instrumentation) → node agent :4318 (OTLP) → Owlpane ingest → console Runtimes

Enable auto-instrumentationCopy Anchor Link

Requires the OpenTelemetry Operator in the cluster.

In Helm values:

autoInstrumentation:
  enabled: true
  samplerRatio: "1"
nodeAgent:
  enabled: true
  otlpHostPort: true

The chart creates an Instrumentation named owlpane that sets:

• OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
• OTEL_METRICS_EXPORTER=otlp (required for runtime vitals/charts)
• OTEL_EXPORTER_OTLP_METRICS_PROTOCOL=http/protobuf

Install/upgrade with the Helm agent OCI chart.

Opt-in on a DeploymentCopy Anchor Link

metadata:
  annotations:
    instrumentation.opentelemetry.io/inject-java: "owlpane/owlpane"

Replace inject-java with inject-nodejs, inject-python, inject-dotnet, or inject-go. The namespace prefix must match the Helm release namespace (default owlpane).

Resource attributesCopy Anchor Link

Ensure pods set or inherit:

• deployment.environment.name
• telemetry.sdk.language (set by language instrumentation)
• service.name (OTEL_SERVICE_NAME or Kubernetes metadata)

What the console showsCopy Anchor Link

Per language: RED cards, slowest operations, top errors, dependencies, workload split series, and runtime metric charts when OTLP metrics are present.

See also SDK runtime telemetry for application-side checklists.

Host / VM workloadsCopy Anchor Link

For processes outside Kubernetes, export OTLP to Owlpane ingest or a standalone collector — Host collector.