← Back to guidesGuide

Prevent OpenTelemetry Collector tail sampling OOMKills without losing incident traces

A practical guide to sizing decision_wait, num_traces, memory_limiter, and exporter queues before the Collector kills your own evidence.

Learn how to control OpenTelemetry Collector tail sampling memory pressure without losing the traces you need during an incident.

Created: July 23, 2026

Published: July 23, 2026

Estimated time45 min

LevelAdvanced

Before you startOpenTelemetry Collector deployed on Kubernetes or an equivalent environment

PlatformsLinux

WhatsApp X LinkedIn

Linux

Flow using kubectl, curl, and promtool to measure memory pressure, tune tail sampling, and promote through a canary.

kubectlcurlpromtooljq

Measure current Collector memory

kubectl -n observability top pod -l app=otel-collector

Extract internal tail sampling metrics

kubectl -n observability port-forward deploy/otel-collector 8888:8888 &
curl -s localhost:8888/metrics | egrep "tail_sampling|memory_limiter|exporter_queue|send_failed"

Validate rules before rollout

promtool check rules otel-tail-sampling-alerts.yaml && promtool test rules otel-tail-sampling.test.yaml

Also available inEvitar OOMKills en tail sampling de OpenTelemetry Collector sin perder trazas de incidente

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.