In previous versions of Grafana, you could only use the API for provisioning data sources and dashboards. But that required the service to be running before you started creating dashboards and you also needed to set up credentials for the HTTP API. In v5.0 we decided to improve this experience by adding a new active provisioning system that uses config files. This will make GitOps more natural as data sources and dashboards can be defined via files that can be version controlled. We hope to extend this system to later add support for users, orgs and alerts as well.
Currently we do not provide any scripts/manifests for configuring Grafana. Rather than spending time learning and creating scripts/manifests for each tool, we think our time is better spent making Grafana easier to provision. Therefore, we heavily relay on the expertise of the community.
It's possible to manage datasources in Grafana by adding one or more yaml config files in the [`provisioning/datasources`](/installation/configuration/#provisioning) directory. Each config file can contain a list of `datasources` that will be added or updated during start up. If the datasource already exists, Grafana will update it to match the configuration file. The config file can also contain a list of datasources that should be deleted. That list is called `deleteDatasources`. Grafana will delete datasources listed in `deleteDatasources` before inserting/updating those in the `datasource` list.
If you are running multiple instances of Grafana you might run into problems if they have different versions of the `datasource.yaml` configuration file. The best way to solve this problem is to add a version number to each datasource in the configuration and increase it when you update the config. Grafana will only update datasources with the same or lower version number than specified in the config. That way, old configs cannot overwrite newer configs if they restart at the same time.
Since not all datasources have the same configuration settings we only have the most common ones as fields. The rest should be stored as a json blob in the `jsonData` field. Here are the most common settings that the core datasources use.
| timeInterval | string | Prometheus, Elasticsearch, InfluxDB, MySQL, PostgreSQL and MSSQL | Lowest interval/step value that should be used for this data source |
Secure json data is a map of settings that will be encrypted with [secret key]({{< relref "../installation/configuration/#secret-key" >}}) from the Grafana config. The purpose of this is only to hide content from the users of the application. This should be used for storing TLS Cert and password that Grafana will append to the request on the server side. All of these settings are optional.
It's possible to manage dashboards in Grafana by adding one or more yaml config files in the [`provisioning/dashboards`](/installation/configuration/#provisioning) directory. Each config file can contain a list of `dashboards providers` that will load dashboards into Grafana from the local filesystem.
When Grafana starts, it will update/insert all dashboards available in the configured path. Then later on poll that path every **updateIntervalSeconds** and look for updated json files and update/insert those into the database.
It's possible to make changes to a provisioned dashboard in the Grafana UI. However, it is not possible to automatically save the changes back to the provisioning source.
If `allowUiUpdates` is set to `true` and you make changes to a provisioned dashboard, you can `Save` the dashboard then changes will be persisted to the Grafana database.
> If a provisioned dashboard is saved from the UI and then later updated from the source, the dashboard stored in the database will always be overwritten. The `version` property in the JSON file will not affect this, even if it is lower than the existing dashboard.
> If a provisioned dashboard is saved from the UI and the source is removed, the dashboard stored in the database will be deleted unless the configuration option `disableDeletion` is set to true.
If `allowUiUpdates` is configured to `false`, you are not able to make changes to a provisioned dashboard. When you click `Save`, Grafana brings up a *Cannot save provisioned dashboard* dialog. The screenshot below illustrates this behavior.
Grafana offers options to export the JSON definition of a dashboard. Either `Copy JSON to Clipboard` or `Save JSON to file` can help you synchronize your dashboard changes back to the provisioning source.
Note: The JSON definition in the input field when using `Copy JSON to Clipboard` or `Save JSON to file` will have the `id` field automatically removed to aid the provisioning workflow.
If the dashboard in the json file contains an [uid](/reference/dashboard/#json-fields), Grafana will force insert/update on that uid. This allows you to migrate dashboards betweens Grafana instances and provisioning Grafana from configuration without breaking the URLs given since the new dashboard URL uses the uid as identifier.
When Grafana starts, it will update/insert all dashboards available in the configured folders. If you modify the file, the dashboard will also be updated.
Alert Notification Channels can be provisioned by adding one or more yaml config files in the [`provisioning/notifiers`](/installation/configuration/#provisioning) directory.
-`notifiers`, a list of alert notifications that will be added or updated during start up. If the notification channel already exists, Grafana will update it to match the configuration file.
By default, exporting a dashboard as JSON will use a sequential identifier to refer to alert notifications. The field `uid` can be optionally specified to specify a string identifier for the alert name.
