Unable to reference data source in alert rule

[siwyroot]

Hello,

I'm trying to deploy GrafanaAlertRuleGroup with operator but when I set datasourceUid to name od the data source rather then UID itself I get always an error that data source is not found. When I set datasourceUid to actual UID everything works but this is not ideal when you want to apply config via gitops. Could not find any other example in docs other then this: https://github.com/grafana/grafana-operator/blob/master/examples/alertrulegroups/resources.yaml#L57

Error
Failed to build evaluator for queries and expressions: failed to build query 'A': data source not found

Datasource

apiVersion: grafana.integreatly.org/v1beta1
kind: GrafanaDatasource
metadata:
  name: ocp-cluster
  labels:
    kustomize.toolkit.fluxcd.io/substitute: disabled
spec:
  instanceSelector:
    matchLabels:
      dashboards: grafana
  valuesFrom:
    - targetPath: "secureJsonData.httpHeaderValue1"
      valueFrom:
        secretKeyRef:
          name: "grafana-sa"
          key: "token"
  datasource:
    access: proxy
    isDefault: true
    jsonData:
      httpHeaderName1: 'Authorization'
      timeInterval: 5s
      tlsSkipVerify: true
    name: ocp-cluster
    type: prometheus
    url: 'https://thanos-querier.openshift-monitoring.svc.cluster.local:9091'
    secureJsonData:
      "httpHeaderValue1": "Bearer ${token}"
  plugins:
    - name: grafana-clock-panel
      version: 1.3.0

Alert Rule Group

---
apiVersion: grafana.integreatly.org/v1beta1
kind: GrafanaAlertRuleGroup
metadata:
  name: monitoring-rules-ingestions
spec:
  folderRef: some-folder-ref
  instanceSelector:
    matchLabels:
      dashboards: "grafana"
  interval: 5m
  rules:
    - uid: dekj368zvsz5sf
      title: 'what ever.'
      condition: C
      data:
        - refId: A
          relativeTimeRange:
            from: 600
            to: 0
          datasourceUid: ocp-cluster
          model:
            datasource:
              type: prometheus
              uid: ocp-cluster
            editorMode: code
            expr: some_expression (hidden)
            instant: true
            intervalMs: 1000
            legendFormat: __auto
            maxDataPoints: 43200
            range: false
            refId: A

[Baarsgaard]

The GrafanaDatasource CR supports setting the UID.
There's two fields for it, the spec.uid is an immutable alternative to the spec.datasource.uid

apiVersion: grafana.integreatly.org/v1beta1
kind: GrafanaDatasource
metadata:
  name: ocp-cluster
spec:
  uid: ocp-cluster       # Overwrites spec.datasource.uid but is immutable.
  datasource:
    name: ocp-cluster
    uid: ocp-cluster     # This is mutable.

I personally prefer to statically define UIDs to ensure the exact same resource is created every time.

[siwyroot]

@Baarsgaard Thanks for that info, I will try it out. On the other hand I could not find in examples this approche (should this be a best practice?) . Dashboard can reference data source by :

spec:
  datasources:
    - inputName: "DS_PROMETHEUS"
      datasourceName: "prom2"

Should GrafanaAlertRuleGroup have the same thing?

[theSuess]

The functionality you refer to here was implemented at a time in which UID management was not yet possible. With UIDs, the intent is clearer and the implementation logic simpler. We're probably not going to add this to alert rule groups but stay with UIDs as the recommendation here

[siwyroot]

Thanks for explanation and solution!