OpenShift Container Platform 4.21 and 4.22 introduce two major node resource management changes. First, AutoSizingReserved is now enabled by default for new clusters in 4.21, automatically calculating CPU and memory reservations for system daemons based on node capacity using tiered formulas. Second, system-reserved-compressible enforcement arrives in 4.22, using cgroup v2 controls to enforce CPU limits on system daemons via the systemd system.slice cgroup. Existing clusters upgrading from 4.20 are protected by an automatically applied MachineConfig that keeps auto-sizing disabled until administrators explicitly opt in. Detailed memory and CPU reservation formulas are provided, along with verification commands and compatibility notes for performance profiles.

13m read timeFrom developers.redhat.com
Post cover image
Table of contents
What's changingThe problem these changes solveWhen these changes take effectHow to calculate the reservationCPU enforcement for system daemonsHow to enable auto-sizing after upgradingVerify the configurationFinal thoughts
288 Impressions