Observability Guides
EspañolEnglish
← Back to guidesGuide

Reducing Prometheus cardinality spikes without breaking alerts

Practical steps to find explosive series, limit labels and validate impact in production

A hands-on guide to detect high-cardinality sources, apply safe relabeling and rollups, and confirm critical alerts remain effective.

Created: April 11, 2026

Published: April 11, 2026

Estimated time60 min
LevelAdvanced
Before you startAccess to Prometheus (web UI and API)
PlatformsDocker / Linux
WhatsAppXLinkedIn

Linux

Commands for Linux hosts running Prometheus under systemd or as a binary: validate configs and inspect metrics.

SSH access to Prometheus hostpromtool installed for config validation
Validate prometheus.yml locally
promtool check config /etc/prometheus/prometheus.yml
Count series via API
curl -sS 'http://127.0.0.1:9090/api/v1/series?match[]=http_requests_total' | jq 'length'
Show Prometheus internal metrics relevant to cardinality
curl -sS 'http://127.0.0.1:9090/metrics' | egrep 'prometheus_tsdb_head_series|prometheus_tsdb_head_chunks|prometheus_engine_queries'
Trigger config reload (graceful)
systemctl kill -s SIGHUP prometheus || curl -X POST http://127.0.0.1:9090/-/reload
Also available inReducir picos de cardinalidad en Prometheus sin romper alertas

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.