Memory pressure triggers random pod eviction in AKS clusters

When AKS cluster experiences memory pressure, Kubernetes kills low priority pods randomly to free memory. This occurs even when cluster/nodepool autoscaling is enabled, as memory release happens faster than scaling. Memory is not compressible like CPU, making exhaustion catastrophic.

Ensure sufficient worker nodes for workload. Monitor for memory leaks and consumption peaks. Use Kubernetes tools for pod priority/preemption, pod disruption budgets, and resource management. Define resource requests and limits for containers. Prevent memory exhaustion as running out is NOT an option.