NOTE
6.2 Linux Control Groups (cgroups)
cgroup v2 resource accounting and control for CPU, memory, I/O, process counts, pressure, and container workloads.
This is a historical learning note and may contain outdated or incomplete understanding.
1. What Is a cgroup?
Linux control groups organize processes into a hierarchy for resource accounting and control.
A useful distinction is:
- namespaces limit what a process can see;
- cgroups limit/account what a group of processes can use.
2. cgroup v2
Modern Linux distributions generally use the unified cgroup v2 hierarchy. Older cgroup v1 examples use separate controller hierarchies and files such as cpu.cfs_quota_us; those examples should be treated as historical rather than copied directly to a modern host.
A cgroup v2 hierarchy is commonly mounted at:
/sys/fs/cgroupProcesses are associated with cgroups through files such as cgroup.procs.
3. Major Controllers
CPU
Controls CPU bandwidth/weight. Common v2 interfaces include cpu.max, cpu.weight, and CPU statistics.
Memory
Tracks and constrains memory usage with files such as memory.current, memory.max, and related pressure/reclaim controls.
I/O
Controls block-device I/O weights or limits and reports per-device statistics.
PIDs
Constrains the number of tasks/processes through the pids controller, helping prevent fork bombs inside a workload.
4. Containers
Container runtimes typically place a container’s processes into dedicated cgroups. Namespaces isolate the container’s view, while cgroups provide accounting and limits.
The host kernel still owns all resources; cgroups do not create a separate kernel.
5. /proc and Observability
Older container environments sometimes exposed host-wide /proc values that ignored cgroup limits, which confused tools such as top and memory-sizing logic.
Modern runtimes and libraries are increasingly cgroup-aware, but applications that size thread pools, heaps, or caches should still verify whether they read host capacity or effective cgroup limits.
6. Pressure and OOM
cgroup v2 supports memory/CPU/I/O pressure accounting and per-cgroup OOM behavior. Resource limits are therefore not only quota mechanisms; they are also important observability and failure-containment boundaries.