What Is Runner Provisioning? From Trigger to Ready Machine
Runner provisioning is everything that happens between a job being triggered and a runner being ready to execute it - launch, boot, register, prepare.
Provisioning is the setup tax of CI. For warm runners it is already paid; for cold ones it is the delay before your first step. Knowing the steps tells you where provisioning time goes and how to cut it.
The provisioning steps
- Allocate compute - launch a VM, container, or pod.
- Boot the OS and start services from the runner image.
- Register the runner with the CI service (token or JIT config).
- Prepare the workspace and begin the job.
Where the time goes
The biggest costs are instance launch and OS boot, especially with large images. Registration is usually fast. Dependency setup inside the job is technically post-provisioning but contributes to time-to-first-useful-work.
Cold vs warm provisioning
Cold provisioning runs all steps when the job arrives, costing tens of seconds to minutes. Warm provisioning does the slow steps in advance so the job grabs a ready runner in about a second. Warm pools are pre-paid provisioning.
Speeding it up
Smaller images boot faster, warm pools hide the cost entirely, and caching shortens in-job setup. Managed platforms tune all three so provisioning is effectively invisible to the developer.
When provisioning fails
Provisioning can fail at any step: capacity unavailable, image pull errors, or a runner that boots but never registers. A robust platform retries on fresh capacity rather than failing the job, so a transient provisioning hiccup does not look like a flaky build to the developer.
Key takeaways
- Provisioning is creating and preparing a runner to run a job.
- The steps are allocate, boot, register, and prepare.
- Instance launch and OS boot dominate cold-provisioning time.
- Warm pools, smaller images, and caching make it near-instant.