This alert is triggered when a Kubernetes node reaches more than 90% of its pod capacity for more than 2 minutes. It indicates that the node has almost no free allocatable pod slots left.

A node running out of pod capacity can cause:

• New pods failing to schedule on the node
• Workload imbalance across the cluster
• Potential service degradation if no other nodes are available
• Increased latency for scheduling or scaling operations

This alert is marked warning, as it may precede node-level failures or application disruptions.

Check node pod allocation:

kubectl get nodes -o wide
kubectl describe node <NODE_NAME>

Check running pods on the node:

kubectl get pods -o wide --all-namespaces | grep <NODE_NAME>

Check node allocatable pods:

kubectl get node <NODE_NAME> -o jsonpath='{.status.allocatable.pods}'

Check cluster-wide pod distribution:

kubectl get pods --all-namespaces -o wide

• Node is heavily loaded with many pods
• Misconfigured deployments with too many replicas on a single node
• DaemonSets consuming pod slots
• Cluster autoscaler not configured or failing
• Pod anti-affinity rules forcing pods onto fewer nodes

1. Review and redistribute workloads across nodes
2. Scale out the cluster by adding more nodes
3. Remove unnecessary pods or workloads from the node
4. Adjust DaemonSets or affinity/anti-affinity rules
5. Enable or tune Cluster Autoscaler if available

• Escalate if multiple nodes are reaching pod capacity
• Page on-call engineer if workloads fail to schedule and impact production
• Monitor cluster autoscaler or take manual action to add nodes

• KubernetesNodeMemoryPressure
• KubernetesNodeDiskPressure
• KubernetesNodeNotReady
• PodPending

• Grafana → Kubernetes / Node Pod Capacity
• Grafana → Cluster Pod Distribution