There are a lot of metrics available, but let's look at some that have proven useful in the past:
- gitlab_runner_jobs: It is wise to monitor the number of pending jobs:
If this number gets too high, consider adding a higher limit for autoscalable Runners.
- gitlab_autoscaling_options_total: Another interesting metric is the type and number of scaling actions that are performed on the Runner:
The actions created and removed should be about the same (too many created is expensive, and too many removed would signal a problem). ...