Transformation

The transform can be used to convert one check into multiple checks programmatically.

alertmanager-check.yaml
apiVersion: canaries.flanksource.com/v1
kind: Canary
metadata:
  name: alertmanager
spec:
  schedule: "@every 5m"
  alertmanager:
    - url: http://kube-prometheus-stack-alertmanager.monitoring:9093
      name: alertmanager-check
      alerts:
        - .*
      ignore:
        - KubeScheduler.*
      exclude_filters:
        namespace: elastic-system
      transform:
        expr: |
          results.alerts.map(r, {
            'name': r.name + r.fingerprint,
            'namespace': 'namespace' in r.labels ? r.labels.namespace : '',
            'labels': r.labels,
            'icon': 'alert',
            'message': r.message,
            'description': r.message,
          }).toJSON()
      relationships:
        components:
          - name:
              label: pod
            namespace:
              label: namespace
            type:
              value: KubernetesPod
        configs:
          - name:
              label: pod
            namespace:
              label: namespace
            type:
              value: Kubernetes::Pod

In the above example, the check returns multiple alerts from Alertmanager. By default, all those alerts are grouped in a single check.

But if we want to display each alert as its own check in the UI, we can use the transform function for this. The transform function takes a Template as input and the output from the template expected are the checks in JSON format.

For example, if there are 9 different alerts, each alert has its own check that can be managed and interacted with equally.

Transform Result

Field	Description	Scheme
name*	Name of the check to use, the name is used to lookup existing checks to update	string
pass*	Set to true if the health check is passing	boolean
duration	Time in ms for duration of the check	int64
invalid	Set to true if the check is not configured correctly, and can never pass	boolean
deletedAt	The time the check or event logically ended	time.Time
error	An error message to be shown, if non-empty sets pass: false automatically	string
icon		Icon
labels		[map[string]string]
message	Informational message to be shown to users	string
namespace		string
start	The time the check or event logically started (e.g if coming from an external source)	time.Time
data		map[string]interface{}
description		string
detail		interface{}
displayType		string
endpoint		string
metrics	Add custom metrics to be exported	[]Metric
transformDeleteStrategy	How to handle checks that were returned in a previous transformation, but not in the current	MarkHealthy | MarkUnhealthy | Ignore
type		string

Transforms can either update checks in-place or return new checks.

Updating checks in-place

The update a check in place, return a JSON string of TransformResult with the updated fields

Creating new checks

To create a new check, return a JSON string of []TransformResult containing the new checks.