Using Jsonnet

Jsonnet is by far the most powerful use-case for Grizzly. It allows resources to be described programmatically, using all the common patterns programmers are familiar with, e.g. variables, libraries and abstractions.

To understand the Jsonnet language, see the Jsonnet tutorial.

Note: The original approach to using Jsonnet in Grizzly involved hidden elements. For details of this deprecated approach, see Hidden Elements.

Here is an example of rendering Prometheus alert and recording rules with Jsonnet:

{
  local rule(name, namespace, spec) = {
    apiVersion: 'grizzly.grafana.com/v1alpha1',
    kind: 'PrometheusRuleGroup',
    metadata: {
      name: name,
      namespace: namespace,
    },
    spec: spec,
  },

  grizzly_alerts:: {
    rules: [{
      alert: 'PromScrapeFailed',
      expr: 'up != 1',
      'for': '1m',
      labels: {
        severity: 'critical',
      },
      annotations: {
        message: 'Prometheus failed to scrape a target {{ $labels.job }}  / {{ $labels.instance }}',
      },
    }],
  },

  grizzly_records:: {
    rules: [{
      record: 'job:up:sum',
      expr: 'sum by(job) (up)',
    }],
  },

  prometheus_rules: [
    rule('grizzly_alert_rules', 'grizzly_rules', $.grizzly_alerts),
    rule('grizzly_recording_rules', 'grizzly_rules', $.grizzly_records),
  ],
}

In this example, we first define a function rule() for creating resources. Then we define our alert and recording rules separately. Finally, we use the rule function to expose these.

In Jsonnet, :: signifies hidden, that is, elements defined with :: won’t be visible in the output. Thus grizzly_alerts and grizzly_records are both internal to the script, and only see the light of day because they are referenced within prometheus_rules.