OpenTelemetry Collector configuration guide
Expert guidance for configuring and deploying the OpenTelemetry Collector to receive, process, and export telemetry.
Rules
| Rule | Description |
|---|
| receivers | Receivers — OTLP, Prometheus, filelog, hostmetrics |
| exporters | Exporters — OTLP/gRPC to Dash0, debug, authentication |
| processors | Processors — memory limiter, resource detection, ordering, sending queue |
| pipelines | Pipelines — service section, per-signal configuration, connectors |
| deployment | Deployment — agent vs gateway patterns, deployment method selection |
| dash0-operator | Dash0 Kubernetes Operator — automated instrumentation, Collector management, Dash0 export |
| collector-helm-chart | Collector Helm chart — presets, modes, image selection |
| opentelemetry-operator | OpenTelemetry Operator — Collector CRD, auto-instrumentation, sidecar |
| raw-manifests | Raw Kubernetes manifests — DaemonSet, Deployment, RBAC, Docker Compose |
| sampling | Sampling — head, tail, load balancing |
| red-metrics | RED metrics — span-derived request rate, error rate, duration histograms |
Key principles
- Processor ordering matters.
Place
memory_limiter first in every pipeline.
Use the exporter's sending_queue with file_storage instead of the batch processor.
Incorrect ordering causes memory exhaustion or data loss.
- One pipeline per signal type.
Define separate pipelines for traces, metrics, and logs.
Mixing signals in a single pipeline breaks processing and causes runtime errors.
- Every declared component must appear in a pipeline.
The Collector rejects configurations that declare receivers, processors, or exporters not referenced by any pipeline.
- Consistent resource enrichment across pipelines.
Apply processors that enrich resource attributes like
resourcedetection and k8sattributes to every signal pipeline (traces, metrics, and logs), not just one.
If one pipeline enriches telemetry with k8s.namespace.name or host.name but another does not, correlation between signals is compromised by incomplete metadata.
- Memory safety is non-negotiable.
Always configure
memory_limiter in production.
Without it, a burst of telemetry can cause the Collector to OOM and crash.
Quick reference
Official documentation