Fix saturation point for shared runners
Fix saturation point for shared runners
Before this, we were measuring shared
-runner utilization as follows:
gitlab_runner_jobs
/
gitlab_runner_concurrent
While the private
and shared-gitlab-org
shards we were doing:
gitlab_runner_jobs
/
gitlab_runner_limit
Because the labelset for gitlab_runner_concurrent
does not include a
system_id
label, we were not recording for
gitlab:component_saturation:ratio
for the shared
shard.
With this, we're using the same metric for the denominator for all
shards (gitlab_runner_limit
).
We're also making sure that we're including all shards into she
shared_runners
saturation component by only excluding private
and
shared-gitlab-org
shards from that component. This means we'll start
including the saas-linux
shards and the macOS and windows
runners. Because of this, we needed to add the shard
label as a
significant label, so we can distinguish which shard we're predicting
saturation for in dashboards.
This has all been merged into a single file to keep the saturation
point definitions the same for all of these shards. The only thing
that changed for private_runners
and shared_runners_gitlab
is the
wording in the description.