Comparison
Pipeline runnervsSelf-hosted runner
Pipeline runner
the job waits in a queue because every machine that could run it is busy, and nobody thinks of that as capacity.
The machine or container that executes a pipeline job. Runners are a capacity pool with all the usual properties — queue depth, saturation, cost and cold start — and queue time is invisible in most build dashboards, which report only how long the job itself took. They are also the most credential-rich compute in an organisation, since they hold the keys to deploy.
Full entry →Self-hosted runner
the build runs on your own machines so it can reach the private network, and now you own patching and isolating it.
Running pipeline jobs on infrastructure you operate rather than the provider's. It is chosen for network access to private resources, for specialised hardware or for cost at volume, and it moves responsibility for isolation onto you. A persistent runner shared between jobs is a genuine security problem: one job can leave things behind for the next, which is why ephemeral, single-use runners are the safe default.
Full entry →