Assess and troubleshoot cluster health

Select the red numbers displayed within the health status area. This action reveals a list of the unhealthy objects in the corresponding table, providing additional insights into the nature of the problems they are facing. For instance, you might notice a node in the cluster that is actively evicting pods, indicating an issue with ongoing memory pressure.

Select the node from the list.

The Davis AI health status at the top signals the issue with conditions. There's a detailed breakdown of the node resource utilization. Pay attention to the Memory section. If memory usage exceeds the allocated requests, it indicates a potential resource strain. Kubernetes, in an effort to maintain node stability, might begin evicting pods to free memory. This is often a response to pods consuming more memory than available, based on their reserved requests.

To identify which pods have been evicted, refer to the Events section for this node.

Select the relevant line in the Events section and choose Show full value to reveal the full details of the event. In this example, you can see that the pod named parallel-processing-medium was evicted due to resource constraints.

Go to the Workloads list and use the filter bar to search for the parallel-processing-medium workload. Select this workload to display details. In the Workload utilization section, you can quickly spot misconfigurations of resource requests.

Use the Ownership information to identify the team responsible for this workload. Reach out to them to discuss and make necessary adjustments to the workload’s resource configurations. For more information regarding ownership, see Assign ownership teams to monitored entities.