Grafana Dashboard in Helm Chart doesn't work

[maxpain]

Hello. I've deployed the dragonfly-operator Helm chart with the following values:

serviceMonitor:
  enabled: true

grafanaDashboard:
  enabled: true

But getting this error:

[Pothulapati]

@maxpain With Dragonfly, We don't deploy the prometheus. The issue seems to be that the datasource is misconfigured. Can you check for the same?

Thanks

[maxpain]

Yes, the provisioned dashboard doesn't have datasource variable

[Pothulapati]

You should be provisioning a prometheus instance right? if you already have one, add it into grafana and set it as the datasource of the dashboard.

The service monitor only configures prometheus to collect metrics 🤔

[maxpain]

@Pothulapati Yes, I have configured Prometheus (actually VictoriaMetrics, which is much better) for my Grafana installation. All my Grafana dashboards are declaratively provisioned via ConfigMaps. Almost all of them are from Helm charts and are working fine (CNPG, for example).

So, the Dragonfly grafana dashboard from the helm chart should be fixed.

[Pothulapati]

@maxpain Sure!

So, probably a way to pass the prometheus data source (including VictoriaMetrics) as a config through the helm chart?

[maxpain]

@Pothulapati I don't know how exactly, but all other helm charts with built-in grafana dashboards don't require configuring the Prometheus data source via helm values. They work out of the box.

[maxpain]

Here is an example: https://github.com/cloudnative-pg/grafana-dashboards/blob/main/charts/cluster/grafana-dashboard.json#L4 They use DS_PROMETHEUS variable.

[Pothulapati]

Looks like DS_Prometheus is a good answer where the user gets the option to select the datasource at the top.

[risenxxx]

I also encountered a similar problem - no datasource selection and the entire dashboard is empty with errors. I was able to fix it locally by adding datasource selection similar to cloudnative-pg to templating.list:

- {
-   "hide": 2,
-   "name": "DS_PROMETHEUS",
-   "query": "prometheus",
-   "skipUrlSync": true,
-   "type": "constant"
- },
+ {
+   "current": {
+     "selected": false,
+     "text": "Prometheus",
+     "value": "prometheus"
+   },
+   "hide": 0,
+   "includeAll": false,
+   "label": "Datasource",
+   "multi": false,
+   "name": "DS_PROMETHEUS",
+   "options": [],
+   "query": "prometheus",
+   "queryValue": "",
+   "refresh": 1,
+   "regex": "",
+   "skipUrlSync": false,
+   "type": "datasource"
+ },

Can we try testing this solution? I think the problem is that "prometheus" query is hardcoded and all $DS_PROMETHEUS mentions are interpolated with this string. So, grafana doesn't ever try to use first available datasource with prometheus type. VictoriaMetrics users usually don't name their datasources "prometheus". For example, mine is called "VictoriaMetrics".

[Pothulapati]

Sure! @risen228 Can you raise a PR?

[risenxxx]

@Pothulapati done: https://github.com/dragonflydb/dragonfly-operator/pull/279