Creating Grafana Alerts for High Pod Memory Usage

July 23, 2025

The blog post explains how to create a Grafana alert for monitoring pod memory usage:

Key Points

• Use resource requests as the memory threshold

• Query to calculate memory percentage:

    (sum(container_memory_usage_bytes{namespace="<my-namespace>", container="<my-container-name>"}) by (pod) / sum(kube_pod_container_resource_requests{exported_namespace="<my-namespace>", container="<my-container-name>", resource="memory"}) by (pod)) * 100
  

• Set alert threshold at 80% of allocated memory

Rationale

The author chooses to use requests instead of limits because they "want to know when a pod is getting close to the amount of memory I have allocated for it."

Labels: Computers