Imagine your application is slowing down, and you suspect a pod is using excessive CPU or memory. How would you identify the problematic pod across multiple nodes in the cluster?

If I suspect that an application is slowing down because one of the pods is consuming excessive CPU or memory, I start by using the Kubernetes command-line tool kubectl top to monitor resource usage.

First, I run the command to list all pods across all namespaces along with their CPU and memory usage. This helps me quickly identify which pod is using the most resources. If needed, I can filter this output or sort it manually to focus on the high-consuming pods.

Next, I cross-reference the pod name with the node it’s running on. This gives me an idea if the issue is isolated to a specific node or workload.

If I need a deeper view, I check the metrics server output and, if configured, use tools like Prometheus and Grafana to visualize trends over time. These tools help me understand whether the spike is temporary or part of a pattern.

If I find a pod that’s consistently consuming abnormal resources, I’ll look into its logs and events, and possibly describe the pod to check for restarts or throttling. From there, I can take appropriate action—like scaling the pod, optimizing the application code, or increasing resource limits.