Should not be necessary to use scalable loki to access its dashboards

[alexppg]

Describe the bug
Hi.
The "bug" I've detected is in this line https://github.com/grafana/loki/blob/main/production/helm/loki/templates/monitoring/dashboards/configmap-1.yaml#L1.

You're not able to see the dashboards using the loki helm chart if you don't use the scalable option. I understand that we should be able to see them regardless.

That being said, I guess it makes sense if seeing the dashboards is useful when running on single binary option, wich I'm not sure. If not, just close the issue.

Thanks for your time and work!

To Reproduce
Steps to reproduce the behavior:

1. Disable isScalable on loki helm chart
2. Enable the dashboards option
3. Apply

Expected behavior
It should install the dashboards.

[jeschkies]

Hm, that's a good point. @trevorwhitney do you remember why the dashboards were excluded? If so we should document this.

[brennerm]

Is there any update on this? We'd also like to use the dashboards with our non-scalable Loki deployment. :)

[trevorwhitney]

The dashboards are just static copies of the JSON that can be built using the loki-mixin. We would need to add support for single binary mode to the loki-mixin first. Since we don't have that yet, we only included dashboards for scalable mode.

[derekhuizhang]

@trevorwhitney Is there a plan to update loki-mixin soon? The release for the new Loki helm chart states that it provides self-monitoring tools: https://grafana.com/blog/2022/12/19/the-only-helm-chart-you-need-for-grafana-loki-is-here/, but half the dashboards are missing metrics by default because they assume that Loki is running in microservices mode

[trevorwhitney]

half the dashboards are missing metrics by default because they assume that Loki is running in microservices mode

mind clarifying that a bit for me? the dashboards shipped with the helm chart should be assuming scalable mode, and if not that's a bug. the loki-mixin can currently be compiled to produce dashboards that work with either scalable or microservices mode. to clarify, scalable to me means read, write, and backend targets, whereas microservices means each component runs in its own pod (which isn't supported by the helm chart, but is supported by the loki-mixin).

my understanding of the problem here is that the helm chart should produce dashboards that are applicable to either single binary or scalable mode. in order to do that, we need a job that can run jsonnet and compile the dashboards at installation time so it can use values from values.yaml to determine the correct labels to include in the dashboard queries so they work for that configuration.

this functionality is not currently planned, but PRs are very welcome, and should be a good PR for anyone looking to get started contributing to the helm chart!