Metrics

Live CPU and memory columns for pods and workloads, backed by the Kubernetes metrics API.

Overview

Krust displays CPU and memory usage for pods and workloads by integrating with the Kubernetes metrics-server API.

Metrics are meant to answer quick operational questions: which pods are hot, whether a rollout changed resource usage, and whether a workload is close to its requested or limited capacity.

Pod Metrics

CPU and memory usage values are displayed directly in the pod list table columns. Values are fetched from the metrics.k8s.io API.

Because metrics live in the same table as status, restarts, age, and namespace, you can sort or filter without switching to a separate dashboard.

Workload Metrics

Aggregated CPU and memory metrics are available for workload resources:

• Deployments
• StatefulSets
• DaemonSets
• Jobs
• CronJobs

Deployment metrics are resolved through ReplicaSet ownership chains.

That means a Deployment view can still show useful aggregate usage even though the raw metrics are reported at the pod level.

Polling

Metrics are polled on a 30-second interval and refresh automatically in the background.

The goal is table-level situational awareness, not high-frequency time-series monitoring. For detailed historical charts, keep using Prometheus or your observability stack.

Graceful Degradation

If metrics-server is not available in the cluster:

• Metric columns display - instead of values
• No errors or crashes occur
• Krust re-checks metrics-server availability periodically

This keeps resource browsing usable in clusters where metrics-server is missing, restricted, or temporarily unavailable.

Troubleshooting

Requirements

• metrics-server must be installed in the cluster (standard in most managed Kubernetes providers). See the official metrics-server docs.
• No additional configuration is needed in Krust