This is a python implementation of the backend needed by the JSON grafana plugin.
Create an .env file which defines the following variables:
HOST=127.0.0.1
PORT=56733
ES_HOST=127.0.0.1
ES_PORT=9200
ES_INDEX=my_index
ES_USERNAME=root
ES_SECRET=password
Uncomment the lines in app/settings.py which load the .env file, install the requirements in a virtualenv and run the application
$ python3 -m venv .virtualenv
$ source .virtualenv/bin/activate
$ pip install -r app/requirements.txt
$ python app/app.py
The api root is served at http://HOST:PORT/grafana/api/v1
Launch the start.sh script
The api root is served at http://HOST:PORT/grafana/api/v1
It uses flask to implement a JSON API. The following routes are served, (see app.py):
This is just for testing purposes, grafana call it when testing the data source to see if it's up and running.
This endpoint returns a list of available fields. It gets a target parameter which contains the characters digited
by the user when filtering the dropdown input.
This endpoint returns the actual query result, could be a timeseries set of data or a table set of data. It receives the following json:
# example of received request
{
'timezone': 'browser',
'panelId': 2,
'dashboardId': 4,
'range': {
'from': '2019-09-06T02:29:24.953Z',
'to': '2019-09-06T08:29:24.953Z',
'raw': {'from': 'now-6h', 'to': 'now'}
},
'rangeRaw': {'from': 'now-6h', 'to': 'now'},
'interval': '20s',
'intervalMs': 20000,
'targets': [
{'data': None, 'target': 'mi', 'refId': 'A', 'hide': False, 'type': 'timeseries'}
],
'maxDataPoints': 960,
'scopedVars': {'Cippo': {'text': '1m', 'value': '1m'},
'__from': {'text': '1567735232102', 'value': '1567735232102'},
'__to': {'text': '1567756832103', 'value': '1567756832103'},
'__interval': {'text': '20s', 'value': '20s'},
'__interval_ms': {'text': 20000, 'value': 20000}},
'adhocFilters': []
}
TBD
The application architecture provides a simple way to connect new sources to the API interface.
The API interface calls a Source Factory (sources/factory.py) to retrieve an instance of a real source class, such instance depends on a setting configuration (now, in the future could depend on a part of the url).
Every source class should inherit from an AbstractSource class (sources/abstract.py) which defines the class interface.
Every source class should implement the following methods:
(string) test ()(list) search (target:string)(dict) query (range:array, interval:int, targets:array, max_data_points:int, scoped_vars:dict, filters:dict)
In order to add a new source, just create a new class defining such methods, tweak the source factory class to return the right class basing on the SOURCE setting.