grafana_datasource_exchange

Usage

This document describes the expected behavior of any charm exposing a grafana_datasource_exchange endpoint.

The reference implementation for this interface is implemented in the cosl library. Charm developers are free to provide alternative libraries as long as they fulfill the behavioral and schematic requirements described in this document.

Direction

The grafana_datasource_exchange interface implements a symmetrical provider/requirer pattern. Symmetrical, in the sense that the role doesn't matter and the data to be exchanged is the same for the requirer and the provider.

flowchart TD
    Provider -- Datasources --> Requirer
    Requirer -- Datasources --> Provider

Behavior

The requirer and the provider need to adhere to a certain set of criteria to be considered compatible with the interface.

Provider & Requirer

  • Is expected to expose a server implementing the grafana source HTTP API. In other words, it's expected to expose one or more valid grafana datasources.
  • Is expected to register each datasource endpoint (one per unit) with a central grafana application and obtain a Datasource UID for each one of them.
  • Is expected to share via application data, as a json-encoded array (sorted by UID), the following information:
    • for each datasource (which technically will likely mean, for each unit of the application):
      • the datasource UID: an arbitrary string, uniquely identifying the datasource
      • the grafana UID: an arbitrary string uniquely identifying a grafana instance
      • the datasource type: a grafana datasource type name (typically will be [one of the built-in ones](https://grafana.com/docs/grafana/latest/datasources/

Relation Data

[Pydantic model]

Example

application_data: {
  datasources:
    [
      {
        type: tempo,
        uid: 0000-0000-0000-0001,
        grafana_uid: 0000-0000-0000-0002,
      },
      {
        type: prometheus,
        uid: 0000-0000-0000-0003,
        grafana_uid: 0000-0000-0000-0004,
      },
    ]
}

Notes

  • Since this interface is symmetrical, each application will likely have to implement both a requirer and a provider endpoint for it, to avoid having strange constraints on the integration topology.
  • The data that is being exchanged comes in part from the application themselves (the datasource type), but in part from another integration (Grafana assigns the UIDs and communicates them back via the grafana_datasource interface). Since we cannot assume which integration is created first and what the event sequence will look like, this interface cannot commit to a specific event by which the data should be written. Instead, the only guarantee an implementer should give, is that eventually the data will be provided.
  • The 'sorting by UID' feature is required to prevent the databag hash to keep flapping and trigger endless cascades of relation-changed events.

Charms implementing this interface

Providers

Requirers

Other charms using this interface

Providers

Requirers