Grafana helm chart - Starred dashboard list disappears after the grafana pod restart

[doanbutar]

Summary of Issue:
Starred dashboard list disappears after the grafana pod restart. The issue only affected the provisioning dashboard.

The user setup used dashboard provisioner and JSON files to import the dashboards from a GitHub repo using a gitsync side car container.

ENVIRONMENT DETAILS
Platform: EKS
Grafana Version: 11

Investigation:

1. I found a similar issue reported here: Dashboards: Starred dashboards cleared after Grafana POD restart  grafana#80398
2. I couldn't reproduce the issue using a single binary grafana, so this seems only to happen for the K8s deployment.

Steps to reproduce:

1. Deploy Grafana using this helm chart: https://github.com/grafana/helm-charts/tree/main/charts/grafana
2. Provision dashboard
3. Star the provisioned dashboard
4. Restart Grafana POD
5. Check starred dashboards

[FedericoAntoniazzi]

Hi @doanbutar, data persistence in grafana is not enabled by default

helm-charts/charts/grafana/values.yaml

Lines 372 to 378 in 6a386f9

	persistence:
	type: pvc
	enabled: false
	# storageClassName: default
	accessModes:
	- ReadWriteOnce
	size: 10Gi

[massimeddu-sj]

In my case I've the a postgres persistence layer, but the problem still exists. I think this is because we are provisioning the dashboards from ConfigMaps, so at each restart of the PODs the dashboards are recreated and so the starred list became empty.

Any idea on how keep the same dashboards internal ids while reloading?

Each ConfigMap dashboard has already an unique uid.

Grafana: 11.1.4
Helm Chart: 8.4.5

[davordbetter]

Same issue here.
I'm using PVC enabled. But still every time pod is recreated (new version) link to default or started dashboards are gone (invalid).

Grafana: 11.2.0
Helm chart: 8.5.1

[FedericoAntoniazzi]

Hi @doanbutar, data persistence in grafana is not enabled by default

helm-charts/charts/grafana/values.yaml

Lines 372 to 378 in 6a386f9

	persistence:
	type: pvc
	enabled: false
	# storageClassName: default
	accessModes:
	- ReadWriteOnce
	size: 10Gi

[massimeddu-sj]

In my case I've the a postgres persistence layer, but the problem still exists. I think this is because we are provisioning the dashboards from ConfigMaps, so at each restart of the PODs the dashboards are recreated and so the starred list became empty.

Any idea on how keep the same dashboards internal ids while reloading?

Each ConfigMap dashboard has already an unique uid.

Grafana: 11.1.4
Helm Chart: 8.4.5

[davordbetter]

Same issue here.
I'm using PVC enabled. But still every time pod is recreated (new version) link to default or started dashboards are gone (invalid).

Grafana: 11.2.0
Helm chart: 8.5.1