Diagnose OpenTelemetry Collector backpressure before signals are dropped

A practical runbook for catching slow exporters, saturated queues, and drops before incident evidence disappears.

Learn how to read OpenTelemetry Collector internal metrics, isolate slow exporters, test a safe canary, and prove queues drain without hiding critical spans, logs, or metrics.

Created: June 28, 2026

Published: June 28, 2026

Estimated time45 min

LevelAdvanced

Before you startAn OpenTelemetry Collector with internal metrics exposed on :8888 or an equivalent endpoint.

PlatformsLinux

WhatsApp X LinkedIn

Linux

A Linux or Kubernetes flow for diagnosing a Collector with internal metrics, promtool, and a controlled canary.

otelcolcurlpromtoolkubectl if the Collector runs on Kubernetes

Validate the configuration locally

otelcol --config collector.yaml --dry-run

Inspect internal metrics

curl -s http://localhost:8888/metrics | grep -E "otelcol_exporter_queue|otelcol_exporter_send_failed|otelcol_processor_batch"

Query queue pressure in Prometheus

promtool query instant http://prometheus:9090 "max(otelcol_exporter_queue_size / otelcol_exporter_queue_capacity) by (exporter)"

Also available inDiagnosticar backpressure en OpenTelemetry Collector antes de perder señales

Content locked

This guide requires both steps before full content is available.

○Click “Like” on this guide.
○Share on WhatsApp, X, LinkedIn, or copy the link.

Access is automatically unlocked as soon as both steps are completed.