Monitoring Kubernetes Namespace

As mentioned earlier, eG Enterprise provides a specialized model for monitoring Kubernetes Namespace (see Figure 1).

Figure 1 : Layer model of Kubernetes Namespace

Each layer of Figure 1 is mapped to tests that report a wide variety of status metrics - e.g., service status, Pod status, Daemonset status, etc. - thus bringing abnormalities to the attention of administrators. Using these metrics, administrators can find quick and accurate answers to the following performance queries:

• Are any nodes running Daemonsets they should not?

• Are any nodes not running the Daemonsets they should?

• Are all Deployments healthy? If not, which are the Deployments that failed to create the desired number of Pod replicas?

• Is any Deployment unavailable?

• Has any Deployment failed to update all desired Pod replicas with changes to the Pod template?

• Are backoff conditions not allowing any Horizonal Pod autoscaler to perform scaling?

• Is any autoscaler unable to compute scales? If so, why?

• Has the scaling ability of any autoscaler been inhibited by the replica limits set?

• Has the target utilization level for scaling been set correctly for all autoscalers?

• Did any autoscaler fail to scale the current replica count to the desired levels?

• How many Pods in a namespace are ready to serve requests? Which ones are they?

• Which Pod is in what phase of its lifecycle?

• Are there any Pods with containers that are not ready to service requests?

• Which Pods are not yet scheduled to nodes, and why?

• Does any Pod have containers that terminated abnormally? If so, which containers and which Pod terminated so, and why?

• Are any Services in a namespace in a Pending state currently? If so, why?