Use data in Grafana

use Horreum to create charts in Grafana

Prerequisites:

  1. Horreum is running, and you have an access token
  2. Grafana is running, and you can update dashboards

Grafana setup

This tutorial uses the JSON API plugin to retrieve data from Horreum. The JSON API Plugin can be installed from the grafana administrator panel.

Install JSON API Grafana Plugin

Next we need to create a datasource that will connect to the Horreum URL. For the example we are connecting to a local development instance of horreum running on localhost using port 8080. The datasource url does not have to include the /api but including it in the datasource saves us from having to include /api each time we fetch data from the datasource.

Create JSON API datasource

That’s it, Grafana is ready to start consuming data from Horreum. The next steps depend on what we want to display in Grafana.

Key Metrics in Grafana

Horreum accepts any json format for a run and uses Labels (TODO link) to extract or calculate key metrics. Metric values, referred to as Label Values in horreum, are automatically calculated for each upload. There is an API to retrieve the values for a specific upload but that would be tedious for comparing across runs. The test api endpoint has a /labelValues (TODO link to documentation) that can retrieve a filtered list of all the label values from each upload.

/test/${testId}/labelValues

The response is a list of json objects that contain the labelValue for each upload. There is actually a json object per dataset but not all runs are parsed into multiple datasets so the distinction is not necessary for most use cases.

[
    {  runId: 101, datasetid: 100, values: {metricName: metricValue,...},
    ...
]

The key data is in the values field of each entry. Add a panel to a dashboard in grafana then select the new JSON API datasource as the datasource for the panel.

Set panel datasource

Grafana can now access the data but we need to define fields to tell the panel how to extract information from the json. There are 3 fields for the /labelValues endpoint: values - the content of the labels runId - the unique ID assigned to the run that was uploaded to horreum datasetId - the unique ID that was assigned to the dataset extracted from the run

The example below defines fields in Grafana for all three but the values field is where the metrics are located.

Define fields for the json

The next step is to define a Grafana transform to turn the values into the dataframe structure Grafana expects. Use the extract fields transform on the values field from the previous step.

Define fields for the json

Grafana should now recognize all the labels as different datapoints. At this point, customizing the grafana panel depends on what values are found in each label and what kind of panel you want to create.

Filtering results

There is a good chance you only want data from certain runs. The /labelValues endpoint supports a filter query parameter to filter out datasets. There are 2 ways to filter:

  1. provide a json object that must exist in the label values.

    For example, if version and txRate are label names then {"version":"1.2.3"} will only include labelValues where version=1.2.3 and {"version":"1.2.3","txRate":2000} will add the txRate=2000 requirement.

curl –query-param “filter={"version":"1.2.3","txRate":2000}” :/api/test/{id}/labelValues

  1. provide a json path (an extractor path from labels) that needs to evaluate to true

    For example, if count is a label we can pass in $.count ? (@ > 10 && @ < 20) to only include datasets where count is between 10 and 20.

curl –query-param “filter="$.count ? (@ > 10 && @ < 20)"” :/api/test/{id}/labelValues

We set the filter parameter by editing the Query for the grafana panel but it will depend .

Define filter for query