Receive daily AI-curated summaries of engineering articles from top tech companies worldwide.
Endigest AI Core Summary
This post announces an improved conversion formula from cgroup v1 CPU shares to cgroup v2 CPU weight for Kubernetes workloads.
•The old linear formula converted 1 CPU (1024m) to only ~39 weight in cgroup v2, far below the default of 100, reducing Kubernetes workload priority against system processes.
•Small CPU requests (e.g., 100m) resulted in cpu.weight=4, too low for sub-cgroup configuration and fine-grained resource distribution.
•The new quadratic formula maps three key anchor points: (2,1), (1024,100), and (262144,10000), yielding cpu.weight=102 for 1 CPU and cpu.weight=17 for 100m.
•The change is implemented at the OCI runtime layer, not in Kubernetes itself; runc 1.3.2 and crun 1.23 have adopted the new formula.
•Existing monitoring tools or custom resource management systems relying on the old formula may need updates.
This summary was automatically generated by AI based on the original article and may not be fully accurate.
Kubernetes
Google Cloud
AWS