Dashboards have confusing panels and dropdowns

[lucabello]

Bug Description

Issue

Field reported that the Loki dashboard has a few issues:

• the Messages input panel shows no data, even though the metric itself is present in Prometheus (maybe a $__interval -> $__rate_interval issue?)
• the dropdown for "Juju Model" shows models where Loki is not deployed; it's not clear that it only makes sense to select models where loki is deployed. Should we use a loki-specific metric to populate the dashboard variables? Or maybe add a juju_charm=loki-k8s?
• the Messages total panel shows no data when selecting the model loki is deployed in, but does show data when the dropdown is set to "All"
• the Index row is empty; it seems to be using cortex_ metrics, so it will probably always be empty for non-HA Loki

Possible Solution

• change $__interval to $__rate_interval where it makes sense (everywhere?)
• select the dropdown variables only for models where Loki is present (not sure if only the juju_model variable or all of them)
• investigate the Messages total dashboard
• either remove the Index row completely, or specify it's for HA

To Reproduce

juju deploy loki-k8s loki
juju deploy prometheus-k8s prometheus
juju deploy grafana-k8s grafana

juju relate loki:metrics-endpoint prometheus
juju relate promethteus:grafana-source grafana
juju relate loki:grafana-dashboards grafana

Environment

-

Relevant log output

-

Additional context

No response

[lucabello]

The Loki dashboard could use a full rewrite. However, until we have time for that, I'll address the most pressing issues only.