Kubernetes 1.37 Memory QoS Is Beta: What cgroup v2 Changes for Memory-Sensitive Workloads

Kubernetes 1.37 Memory QoS Is Beta: What cgroup v2 Changes for Memory-Sensitive Workloads

Kubernetes 1.37 promotes Memory QoS to Beta and enables the feature gate by default. On Linux cgroup v2 nodes, kubelet can use memory.min, memory.low and memory.high to express protection and throttling. The default kubelet configuration does not automatically turn on memory throttling or tiered res

Updated September 14, 2026. Kubernetes 1.37 promotes Memory QoS to Beta and enables the feature gate by default. On Linux cgroup v2 nodes, kubelet can use memory.min, memory.low and memory.high to express protection and throttling. The default kubelet configuration does not automatically turn on memory throttling or tiered reservation.

The distinction between 'feature gate enabled' and 'policy configured' matters. Operators should not assume an upgrade silently gives every pod hard memory guarantees. Tiered reservation is node-wide, and page cache charged to a container can also be protected, so policy choices can change reclaim behavior between neighboring workloads.

A practical review checklist

  • Confirm nodes actually use cgroup v2.
  • Review kubelet memoryThrottlingFactor and memoryReservationPolicy.
  • Model mixed Guaranteed and Burstable workloads before enabling tiered reservation.
  • Watch node pressure, reclaim and workload latency after rollout.
  • Use canary node pools rather than changing every node at once.

Sources and verification

Primary source: Kubernetes Blog — “Kubernetes v1.37: Memory QoS Graduates to Beta” (September 14, 2026).

This article distinguishes confirmed release facts from engineering interpretation. Dynamic details such as support status, limits and availability should be rechecked against the primary documentation before a production change.