← Back to guidesGuide

Fix Kubernetes memory alerts without falling for the page cache trap

A practical Prometheus workflow so memory pages fire on RSS, OOMs, and real pressure, not because container_memory_usage_bytes wanted attention.

Learn how to redesign Kubernetes memory saturation alerts by comparing usage, working set, RSS, cache, limits, OOMKilled events, and promtool validation.

Created: August 26, 2026

Published: August 26, 2026

Estimated time40 min

LevelIntermediate

Before you startPrometheus scraping kubelet/cAdvisor and kube-state-metrics

PlatformsLinux / Docker

WhatsApp X LinkedIn

Linux

Diagnose from a Linux workstation with kubectl, the Prometheus API, and promtool.

kubectlcurlpromtool

Compare memory as Kubernetes sees it

kubectl top pod -n shop

Inspect OOM events

kubectl describe pod -n shop checkout-api-7d9f | egrep -i 'oom|killed|memory|restart'

Query working set

curl -G 'http://localhost:9090/api/v1/query' --data-urlencode 'query=sum by (pod,container)(container_memory_working_set_bytes{namespace="shop",container!=""})'

Test rules

promtool test rules tests/k8s-memory-alerts.test.yaml

Also available inCorregir alertas de memoria en Kubernetes sin caer en la trampa de page cache

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.