Skip to content

Datasource#

Choose a scope#

There are three different scopes in which you can define a datasource, depending on how much you want it to be shared.

  • for common use cases, use higher scopes to reuse the same datasource on multiple dashboards
  • for more specific needs, use lower scopes to restrict the datasource availability to a specific set of (or even a single) dashboard(s).

Dashboard level#

We have extended the dashboard spec to include a list of datasources:

interface DashboardSpec {
  // ... existing dashboard spec ...
  datasources: Record<string, DatasourceSpec>;
}

Of course, the scope of such definition is the dashboard where it is defined. It cannot be used outside the dashboard.

Project level#

In case you would like to share a datasource across different dashboards in the same project, you will need to create a Datasource.

kind: "Datasource"
metadata:
  name: <string>
  project: <string>
spec: <Datasource specification>

Global level#

When we talk about scope and user permission in a REST API, the easiest way is to associate one permission per endpoint. If we want to provide a datasource shared by all projects, then it makes sense to have a different object that is living outside a project.

That’s why we have another resource called GlobalDatasource

kind: "GlobalDatasource"
metadata:
  name: <string>
spec: <Datasource specification>

Datasource specification#

display: <Display specification> # Optional

# If true, then it's the default datasource for the type defined in the plugin.
default: <boolean> | default = false # Optional

# The definition of the plugin datasource
plugin: <plugin>

Plugin definition#

# The type of the datasource. For example, `PrometheusDatasource`
kind: <string>

# The actual definition of the datasource. It will depend on the type defined in the previous field `kind`
spec: <Plugin specification>

We are supporting only prometheus as a datasource for the moment. Please look at the documentation to know the spec for the Prometheus datasource.

Selecting / Referencing a Datasource#

In the panels, you will be able to select a datasource. Like proposed in the first draft, the selector will be like this:

interface DatasourceSelector {
  kind: string;
  // name is the unique name of the datasource to use
  // If name is omitted, that effectively means "use the default datasource for this kind".
  name?: string;
}

The priority order is "Local datasource in the dashboard" > "Project datasource" > "Global datasource".

So if by any chance you have a local datasource that is named exactly like a Project datasource, or a Global datasource, we will consider that the user intentionally wanted to override the upper datasource. We will use the one with the smallest scope.

How to use the Perses' proxy#

As described before, you can provide a proxy configuration that will be used by the Perses server to redirect any queries to the datasource.

It means in the case of the Prometheus datasource, if the field directUrl is not set then the FE needs to use the Perses server to contact the datasource.

For that, the FE will have to determinate which URL should be used to contact the Perses server based on what kind of datasource is used.

  • datasource is at project scope.
  var datasource; 
  if datasource.kind == 'Datasource'; then 
    url= '/proxy/projects' + datasource.metadata.project + '/datasources/' + datasource.metadata.name
  • datasource is at global scope.
  var datasource; 
  if datasource.kind == 'GlobalDatasource'; then 
    url= '/proxy/globaldatasources/' + datasource.metadata.name

API definition#

Datasource#

Get a list of Datasource#

GET /api/v1/projects/<project_name>/datasources

URL query parameters:

  • kind = <string> : should be used to filter the list of datasources with a specific kind
  • default = <boolean> : should be used to filter the list of datasources to only have the default one. You should have one default datasource per kind
  • name = <string> : should be used to filter the list of datasources based on the prefix name.

Example:

The following query should return an empty list or a list containing a single Prometheus datasource.

GET /api/v1/projects/<project_name>/datasources?kind=Prometheus&default=true

Get a single Datasource#

GET /api/v1/projects/<project_name>/datasources/<datasource_name>

Create a single Datasource#

POST /api/v1/projects/<project_name>/datasources

Update a single Datasource#

PUT /api/v1/projects/<project_name>/datasources/<datasource_name>

Delete a single Datasource#

DELETE /api/v1/projects/<project_name>/datasources/<datasource_name>

GlobalDatasource#

Get a list of GlobalDatasource#

GET /api/v1/globaldatasources

URL query parameters:

  • kind = <string> : should be used to filter the list of datasource with a specific kind
  • default = <boolean> : should be used to filter the list of datasource to only have the default one. You should have one default datasource per kind
  • name = <string> : should be used to filter the list of datasource based on the prefix name.

Example:

The following query should return an empty list or a list containing a single Prometheus datasource.

GET /api/v1/globaldatasources?kind=Prometheus&default=true

Get a single GlobalDatasource#

GET /api/v1/globaldatasources/<name>

Create a single GlobalDatasource#

POST /api/v1/globaldatasources

Update a single GlobalDatasource#

PUT /api/v1/globaldatasources/<name>

Delete a single GlobalDatasource#

DELETE /api/v1/globaldatasources/<name>