AWS rate limit causes runner to continuously try to spawn new instances
From a customer:
I suspect that the GitLab runner process uses the existing number of machines to decide if it needs to spawn new machines. if there is an error, it spawns a new one for good measure. what happens is that under a rate limit scenario, the query for the number of active instances always returns an error. this causes the runner to spawn another instance only for the scenario to repeat and create more instances.
it seems like it happens when we use Terraform with AWS.
that should trigger the same condition that we experience and you can see the runners getting spawned incessantly.
They are using GitLab and Runner 10.6 currently.