expose annotations to linter

[diurnalist]

these are structured similarly to the 'templating' field, but in general do not reliably have a 'type' field exposed.

use-case for this is catching when datasources are defined in annotations but they're not derived from dashboard inputs.

[rgeyer]

When would this get used?

The opinion of the dashboard linter is that a dashboard should always have a templated data source, which the user can select.

By using an annotated data source instead, that directly violates the template-datasource-rule.

Or.. Am I missing something?

[diurnalist]

maybe I am misunderstanding, but this is about defining annotations w/in the dashboard. Annotations can be defined by querying a datasource. for example, we had some dashboards at $JOB that had this in the top-level JSON schema:

  "annotations": {
    "list": [
      {
        "datasource": {
          "type": "prometheus",
          "uid": "0c52be2e-c5dd-48b3-b134-99d970fd5bc3"
        },
        "enable": true,
        "expr": "changes(kube_deployment_status_replicas{k8s_namespace_name=\"$kube_namespace\", deployment_environment=\"$env\", deployment_stack=\"$stack\"}[$__interval]) > 0",
        "iconColor": "red",
        "name": "Deployment annotations",
        "titleFormat": "Deployment"
      },
      {
        "builtIn": 1,
        "datasource": {
          "type": "grafana",
          "uid": "-- Grafana --"
        },
        "enable": true,
        "hide": true,
        "iconColor": "rgba(0, 211, 255, 1)",
        "name": "Annotations & Alerts",
        "type": "dashboard"
      }
    ]
  },

because the linter was not looking at .annotations on the model, it didn't detect the fact that this datasource wasn't templated. and yes, the intent is to be able to use the templated datasource rule in more places :) - there are a lot of places where datasources can be configured, i am finding :p

[rgeyer]

Ohh ohh right.. I'm picking up what you're putting down now. I conflated this with the __inputs value that you can use to replace a hard-coded data source when importing a dashboard. Totally different thing, but my brain combined them.

So for this, we'd want to make sure that the data source used by the annotation is the templated data source.

That warrants another rule, as it wouldn't make sense in panel-template-rule, since it isn't really a panel. Maybe a dashboard rule.. 🤔 I'll have to give that some thought.

Can I ask you to create an issue to go along with this, to ultimately implement a rule that takes advantage of the visibility of the annotations? I can get this merged at that point.

[diurnalist]

sure! i filed #197

fwiw we are using this project as part of a custom lint script and I wrote a new rule datasource-from-inputs-rule that will detect and automatically fix any instances where datasources are not templated.