Back to news
cloud Priority 4/5 4/30/2026, 11:05:52 AM

Kubernetes v1.36 Enhances Memory Quality of Service with Tiered Protection Mechanism

Kubernetes v1.36 Enhances Memory Quality of Service with Tiered Protection Mechanism

The latest update to the Memory Quality of Service feature allows Kubernetes to utilize the memory.low and memory.min settings of the cgroup v2 controller. This mechanism ensures that workloads receive tiered protection against reclaim pressure by specifying guaranteed memory levels for individual containers. By defining these thresholds, the kubelet can prevent the kernel from aggressively reclaiming memory from critical processes when the node faces memory pressure.

Related tools

Recommended tools for this topic

These picks prioritize high-intent tools relevant to this topic. Some links may include partner or affiliate tracking.

#kubernetes#devops#cloud#official

Comparison

AspectBefore / AlternativeAfter / This
Cgroup versionLimited support primarily focused on cgroup v1Requirement for cgroup v2 for full Memory QoS
Memory protectionBest-effort management without specific kernel hintsTiered protection using memory.low and memory.min
Reclaim behaviorKernel reclaims memory based on global LRU listsKernel respects defined memory floors during reclaim
Resource isolationNo clear separation of memory availability guaranteesExplicit isolation levels based on workload priority

Action Checklist

  1. Verify Linux nodes are running with cgroup v2 enabled Memory QoS is not compatible with cgroup v1 environments
  2. Enable the MemoryQoS feature gate in the kubelet configuration This feature is currently in alpha status
  3. Configure memoryRequest and memoryLimit for critical pods The kubelet uses these values to calculate cgroup settings
  4. Monitor node-level memory reclaim metrics Ensure that tiered protection is preventing unnecessary OOM events

Source: Kubernetes Blog

This page summarizes the original source. Check the source for full details.

Related