Explain how to populate your metrics dashboard
Problem to solve
The TW for devopsmonitor (@aqualls) has gotten conflicting information on what https://docs.gitlab.com/ee/operations/metrics/#populate-your-metrics-dashboard should contain. This issue documents the feedback she's gotten so she can transfer the question over to the Monitor team for a decision. Since she's not sure which engineer to assign, she's assigning this issue over to @mnohr for distribution to the right person.
Further details
Here's the information gathered so far:
In !36886 (comment 380514089) @mikolaj_wawrzyniak commented on the "Populate your metrics dashboard" subheading:
Thought I wonder if we should outline Auto DevOps configuration process here, this seems as duplication content from https://docs.gitlab.com/ee/topics/autodevops/index.html section. Since devopsmonitor does not own Auto DevOps process it is highly likely that some changes to that process can slip our attention, and than this section will become outdated. Maybe we should only link here to either: https://docs.gitlab.com/ee/topics/autodevops/index.html#enablingdisabling-auto-devops or https://docs.gitlab.com/ee/topics/autodevops/quick_start_guide.html wdyt?
In #229593 (comment 382608900) Amy copied over a statement from @tristan.read in #s_monitor: This might be interesting to y’all - you don’t need to do an actual deployment to get the metrics to appear, it’s enough to go to the ‘Environments’ tab and add an environment called production. This cuts out the need to have a working runner, and means the time taken to go from clean gdk to visible metrics is very short Note: Removing this option based on this comment
In #229593 (comment 382613348) Tristan provided more thoughts:
@aqualls I thinking more that we should document a rapid development environment setup for metrics, probably here: https://gitlab.com/gitlab-org/gitlab-development-kit/-/tree/master/doc/howto/prometheus.
it’s enough to go to the ‘Environments’ tab and add an environment called production
In a real-world situation the environment is often created by the first deployment, and this is when metrics would generally start being populated, so the product docs make sense for a regular user who has followed the AutoDevops setup up to that point. Manually creating an environment before doing a deployment would not be super useful here since without a deployment you wouldn't expect there to be metrics.
Some more detail: @seanarnold created a prometheus project that produces metrics without needing a local kubernetes deployment - see here: https://gitlab.com/seanarnold/local-prometheus-metrics. When trying this out with a clean development environment, I found that you can skip the Connecting to GitLab metrics steps to make setup even faster, by manually adding a
production
environment in the UI.
Proposal
As the technical writer for the stage, Amy can polish what she's given, but shouldn't be the person who blindly chooses one of multiple "getting started" paths, because she doesn't know which one is right.
One of the largest issues Nadia identified in the card sort exercise for Monitor was the importance of the "getting started" documentation. We need to provide a simple, comprehensible path forward for new users that maximizes their chances of success.
We want to balance providing a SSOT with providing a simple experience that has a hope of success.
Who can address the issue
An engineer from devopsmonitor.
Other links/references
- Related to !36886 (merged)
- Related to #229593 (closed) which is the followup from 36886