Self-hosted runner feature comparison
1 month ago
This page offers a comparison of the features available for CircleCI machine runner vs. CircleCI container runner.
Comparison matrix
| Feature | Description | Machine Runner | Container Runner |
|---|---|---|---|
Environment |
Where jobs run and how resources are managed |
Full VM/bare metal |
Kubernetes Pods |
Configuration File |
Primary file for runner setup |
YAML config (location depends on OS) |
Helm |
Resource Control |
How compute resources are allocated |
System level |
Kubernetes limits |
Architecture Support |
CPU architectures available for builds |
AMD64/ARM64 with |
|
Job Isolation |
How jobs are kept separate |
VM-level |
Container-level |
Scaling |
How runner capacity scales with demand |
Kubernetes-based auto-scaling |
|
Service Containers |
Support for additional services in build |
Not supported |
Extensive support with resource controls |
Execution Modes |
How runner handles job execution |
|
Pod-based execution |
Docker Support |
How Docker-in-Docker is handled |
Native OS support |
Possible with third-party tools |
Resource Cleanup |
How build environments are cleaned after jobs |
Working directory cleanup |
Container lifecycle |
Feature Highlights
Machine Runner
-
Latest generation of machine-based runners.
-
Full access to underlying operating system.
-
Ideal for workflows requiring VM-level isolation.
-
Consistent environment variables with
CIRCLECI_RUNNER_*prefix. -
Supports SSH connectivity for debugging.
Container Runner
-
Kubernetes-native solution for container-based workflows.
-
Fine-grained resource allocation and constraints.
-
Supports service container configurations with pattern matching.
-
Advanced node selection via Kubernetes annotations.
-
Custom security contexts and volume mounting.
-
Support for spot instances via tolerations.
-
Configurable resource limits for primary and service containers.
-
Supports SSH connectivity for debugging - Open preview.