What Is a Runner Fleet? Managing Runners at Scale
A runner fleet is the entire collection of runners an organization operates across teams, sizes, and operating systems - and managing it well is a real engineering job.
One runner is trivial. A hundred runners across Linux, Windows, GPU, and large sizes, serving dozens of repos with the right access controls and patch levels, is a fleet - and fleets need real operational discipline.
What a fleet contains
A fleet spans multiple runner pools by OS and size, with labels and groups for routing and access, autoscaling per pool, and a shared image-and-patch baseline. It is the sum of all your CI compute.
What fleet management involves
- Capacity planning and autoscaling across every pool.
- Image building, patching, and version updates fleet-wide.
- Access policy via runner groups and labels.
- Monitoring for offline, stale, or stuck runners.
- Cost tracking and utilization across teams.
Why fleets get hard
Complexity compounds: more OSes, more sizes, more repos, more security boundaries. A self-managed fleet quietly becomes a platform-engineering project with on-call, patch cadence, and scaling code to maintain.
The managed-fleet option
A managed platform operates the entire fleet for you - every pool, image, and scaling decision - so you consume capacity without running the fleet. Latchkey is your runner fleet, minus the fleet ops.
Signs your fleet has outgrown DIY
Recurring runner incidents, a growing backlog of patch and scaling tickets, and an engineer who has quietly become "the runner person" are all signals that operating the fleet now costs more than it saves. That is usually the point teams move to a managed fleet.
Key takeaways
- A runner fleet is all the runners an organization operates.
- It spans pools by OS and size, with labels, groups, and autoscaling.
- Managing a fleet is a platform-engineering job at scale.
- A managed platform runs the whole fleet for you.