Cgroupv2 Is Coming Soon To a Cluster Near You
summary
Kubernetes 1.25 enables cgroupv2 GA and recommends using the systemd cgroup driver with kernel ≥5.8; containerd and runC already support it and SIG‑Node added continuous tests. CPU behavior: requests map to CPU weight (shares) and scheduling, limits use CFS bandwidth (default period 100ms) and can throttle even when spare CPU exists. Memory gains four soft/hard knobs (memory.min/low/high/max) and an alpha MemoryQoS feature (default high = 0.8×limit) to throttle before OOM. A GKE 1.25 demo shows systemd scope units and per‑container cgroup settings.
AI-written summary · about
chapters (10)
- 0:00Start
- 0:39The 10k view - Resources
- 1:54Resource Management Requirements
- 5:15What's new in cgroups v2?
- 6:40cgroups v1 vs v2 hierarchy
- 8:30Journey of Pod Spec
- 12:01CPU Limits - Cfs Bandwidth Control
- 13:32The CPU Limits Debate
- 19:03cgroups v2 testing
- 20:49Test your apps on cgroup v2
related talks
32mWhat is cgroups v2? and how it affects our application
34mKubernetes SIG Node Intro and Deep Dive
23mThis is The Way: A Crash Course on the Intricacies of Managing CPUs...
08mKeynote: Kubernetes Project Updates
47mCPU & Memory, Requests & Limits, QoS & CFS: The JVM in the Kubernetes Pod
36mTaming the Beast: Advanced Resource Management With Kubernetes
23mPractice of Fine-grained Cgroups Resources Scheduling in Kubernetes
21mDo Not Disturb Mode: Fine-Tuning Resources for Latency Sensitive Workloads