# Linux kernel developers propose removing broken RT_GROUP_SCHED scheduler feature

> **Key takeaway:** RT_GROUP_SCHED, a cgroups v1-era real-time group scheduler that all major distributions already disable, is slated for removal from the Linux kernel as unfixable.

**Published:** 2026-11-06T11:09:14+00:00 
**Source:** Phoronix Benchmarks & Drivers 
**Category:** kernel-os 
**Canonical URL:** https://fosswire.org/news/linux-kernel-developers-propose-removing-broken-rtgroupsched.html 

## Summary
Linux kernel developers are moving to remove RT_GROUP_SCHED, a feature that enables group scheduling of real-time tasks through cgroups. The subsystem was designed for cgroups v1 and has become incompatible with the modern v2 hierarchy, preventing proper runtime assignment and blocking standard real-time policy configuration for tasks.

Linutronix engineer Sebastian Andrzej Siewior submitted the removal patch, noting that all major distributions including Debian, openSUSE, Fedora, and Arch Linux already compile their kernels with this feature disabled. The proposed replacement path involves the existing DL fair server for RT throttling and ongoing work on Hierarchical Constant Bandwidth Server for grouping deadline-scheduled tasks.

## What breaks in RT_GROUP_SCHED and what replaces it
- **SCHED_FIFO/SCHED_RR blocking:** When RT_GROUP_SCHED is enabled, tasks cannot have their scheduling policy set to SCHED_FIFO or SCHED_RR because their cgroup lacks assigned runtime, and cgroups v2 provides no mechanism to assign that runtime.
- **Global sysctl interference:** The root cgroup retains all runtime allocation, which prevents reducing the system-wide sched_rt_runtime_us parameter since any reduction would fall below currently available bandwidth.
- **SCHED_DEADLINE admission control failure:** While SCHED_DEADLINE tasks can technically be scheduled, admission control becomes stuck and can only be bypassed by setting sched_rt_runtime_us to -1, disabling the safety mechanism entirely.
- **Distribution consensus:** All examined major distribution kernels compile without CONFIG_RT_GROUP_SCHED, indicating no production dependency on the feature.
- **Replacement architecture:** The DL fair server already handles RT throttling, and HCBS development aims to provide proper grouping of real-time tasks under unified deadline configurations.

## Why it matters
Removal of RT_GROUP_SCHED simplifies the scheduler codebase and eliminates a configuration trap that has silently broken real-time workloads when inadvertently enabled. For embedded and industrial Linux users who require deterministic scheduling, this clears the path toward HCBS as a modern, cgroup-v2-compatible replacement for grouped real-time resource management.