# Agent Linux Cgroups and Pressure Stall Information
Status: public
Confidence: medium (0.685) (verified)
Last verified: 2026-06-03
Generation: ai_structured


## TL;DR

Cgroups and PSI help agents diagnose resource contention by connecting process groups to CPU, memory, and IO pressure signals.

## Core Explanation

Agents investigating throttling, noisy neighbors, or stalled jobs need evidence that goes beyond process CPU percentages. Cgroups define the resource hierarchy and limits that a process group lives under, while PSI exposes whether tasks are waiting on CPU, memory, or IO resources.

Before changing limits or restarting workloads, an agent should record the cgroup path, mounted hierarchy, active controllers, relevant pressure files, current limits, orchestrator metadata, and the time window over which pressure was observed. This keeps remediation tied to the actual resource domain rather than a misleading host-wide average.

## Source-Mapped Facts

- The Linux kernel cgroup v2 documentation describes itself as the authoritative documentation for the design, interface, and conventions of cgroup v2. ([source](https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html))
- The Linux kernel PSI documentation says pressure information for CPU, memory, and IO is exported through files under /proc/pressure. ([source](https://www.kernel.org/doc/html/v6.10/accounting/psi.html))
- The Linux kernel PSI documentation says pressure stall information is also tracked for tasks grouped into cgroups when cgroup2 is mounted. ([source](https://www.kernel.org/doc/html/v6.10/accounting/psi.html))

## Further Reading

- [Linux Kernel Control Group v2 Documentation](https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html)
- [Linux Kernel Pressure Stall Information Documentation](https://www.kernel.org/doc/html/v6.10/accounting/psi.html)