Linux Kernel Control Group v2
summary
Explains that cgroup v2 replaces per-controller trees with a unified hierarchy so controllers can coordinate, and that controllers must be explicitly enabled per subtree via cgroup.subtree_control. Shows exact mechanics: move tasks by echoing PIDs into cgroup.procs, throttle CPU with the new cpu.max (quota/period in μs), and use delegation so unprivileged users can manage subtrees. Notes thread-vs-process semantics, a bypass mode to skip middle controllers, and that distributions still run v1 by default so v1/v2 coexist during migration.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 0:22What is Control Port
- 2:45Process Hierarchy
- 10:35Control Hierarchy
- 16:47Controllers
- 24:25Naming Convention
- 25:55Controller
- 30:53Bypass Mode
- 32:27Control Manager
- 35:04CPU Controller
related talks
57mAn introduction to control groups (cgroups) version 2
01h 02mDiving deeper into control groups (cgroups) v2
27msystemd cgroup delegation and control processes
48mBrave new world of unified cgroup hierarchy
38mUnderstanding and Working with the Cgroups Interface
35mFuture-Proof Your Applications: Avoid Breakdowns With Cgroup Chan...
34mCgroups and Enterprise Users
32mWhat is cgroups v2? and how it affects our application