Docs: Add documentation about runner managers
Problem to solve
We don't have a dedicated section for runner managers, and documentation is fairly limited for this particular runner component/process/entity.
NOTE: There is an ongoing discussion to change runner manager
to runner
, and runner
to runner configuration
(all runners that are linked to the same configuration). #418997. This documentation update shouldn't be blocked by the name update, unless there are some functional considerations along with the name change.
Further details
See runner Glossary for current terminology.
Proposal
After the user reads the docs, they should understand:
What a runner manager is:
- How they are used, in what scenarios (autoscaling, fleeting)/
- How they're different to the entity we call a 'runner'.
- When they are created and when they're destroyed (should include information about stale runner managers being cleaned up after 7 days (Ultimate only).
Runner manager registration and management:
- How to use an authentication token to register the same runner on multiple machines, which is something that is not covered atm.
- Mention that the system ID is retrieved from local machine identifiers. Provide some examples of how to erase them depending on the host platform.
The location for the docs should be decided. Possible docs locations:
- Plan and operate a fleet of shared runners.
- Autoscaling configuration: one or more runner managers, multiple workers.
- This is also a valid concern outside of the realm of autoscaling (if to a lesser degree). Possible it should be documented elsewhere, like in Registering runners.
Who can address the issue
TW for grouprunner with input from engineering.