The GitHub Actions Runner Controller (ARC) project released version 0.14.0, bringing several significant capabilities that improve how organizations manage self-hosted runners at scale on Kubernetes clusters. This release addresses some of the most requested features from teams running hundreds or thousands of self-hosted runners in production environments.
Multi-Label Support for Runner Scale Sets
The headline feature introduces multilabel support for runner scale sets. Previously, ARC runner deployments required separate scale set resources for each unique label combination — a limitation that multiplied Kubernetes resources and complexity as labeling requirements evolved. A team needing runners labeled “gpu”, “linux”, and “large” needed separate scale sets for each combination, even though they were functionally similar workloads.
With multilabel support, a single scale set can now satisfy workflow runs requesting any combination of configured labels. This dramatically simplifies Kubernetes resource management for teams running diverse workload types — GPU-enabled runners, specific architectures, security-hardened environments, and specialized toolchains can now coexist under unified scale set configurations.
The change reduces the operational surface area for platform teams. Fewer custom resources means less YAML to maintain, fewer controller reconciliation loops, and simpler debugging when things go wrong. It also reduces the cognitive load for developer teams submitting workflow jobs — they can rely on standard label combinations without needing to know the underlying ARC topology.
ScaleSet Library Client
ARC 0.14.0 introduces a new ScaleSet library client, providing a programmatic interface for managing runner scale sets directly. This library enables custom automation, monitoring integrations, and platform team tooling that operates against ARC’s core scaling logic without needing to interact with Kubernetes APIs directly.
For organizations building internal platforms around GitHub Actions, the library client opens possibilities for:
- Custom dashboards: Build platform-specific monitoring that shows runner utilization, queue depths, and job distribution across teams
- Predictive scaling: Implement custom autoscalers based on historical patterns, business hours, or calendar events rather than just current queue depth
- IaC integration: Tight integration with existing infrastructure-as-code workflows where scale sets need to be provisioned as part of broader environment setup
The library exposes REST APIs that ARC uses internally, giving platform teams the same primitives the controller has access to for managing runner lifecycles, scaling decisions, and health monitoring.
Experimental Helm Charts
The release includes updated experimental Helm charts that preview future deployment patterns. These charts incorporate lessons from production ARC deployments at scale, offering more flexible configuration options for pod templates, resource constraints, and auto-scaling behavior.
Teams evaluating ARC for the first time or planning migration from the legacy runner architecture should review the experimental charts. They represent the direction the project is heading for stable deployment patterns, including improved security defaults, better resource isolation, and simplified configuration structure.
Operational Considerations and Best Practices
While multilabel support reduces resource sprawl, teams should plan label strategy carefully. Overly broad label assignments can create scheduling challenges where high-demand runners consume capacity needed for specialized workloads. The new library client provides primitives for building visibility into these dynamics, helping platform teams identify contention and optimize their runner pools.
As ARC continues maturing, it’s becoming the de facto standard for running GitHub Actions at scale on Kubernetes — filling a critical gap for organizations with compliance, performance, or cost requirements that preclude using GitHub’s hosted runners. Version 0.14.0 represents another step toward production-ready self-hosted runner management.
Sources
- GitHub Changelog — “Actions Runner Controller version 0.14.0” (March 19, 2026)
- ARC GitHub Repository — Release Notes and Documentation
- ARC 0.14.0 Migration Guide