Site Reliability Guardian

A guardian is the grouping of objectives. It is built around a set of entities reflecting a service or application you want to safeguard.

A guardian provides you with a default automation workflow that performs the objective validation. As a result, a guardian always represents the latest validation result derived from the objectives.

Objectives are means for measuring the performance, availability, capacity, and security of your services. Objectives are measured by indicators. You can define an objective for your guardian that is validated on demand or automatically.

An indicator is a value against which the warning and failure thresholds are checked using a comparison operator. To retrieve an indicator value, use DQL or reference an existing SLO.

The static warning and failure thresholds determine whether the measured value of the indicator meets the objective, is close to violating the objective, or violates the objective.

Warning and failure are optional; objective validation can vary:

Auto-adaptive thresholds are dynamic limits that adjust over time based on previous validations. If an objective changes its behavior, the threshold adapts automatically.

The comparison operator defines whether the objective is met: the indicator is less than or equal to (A lower value is good for my result), or it is greater than or equal to (A higher value is good for my result), the warning and failure threshold.

To organize your guardians, you can assign tags to them. Tags use the key:value format, with the value being optional.

To assign a tag to your guardian, either specify it in the Add tags to your guardian section during guardian creation or add the tag later in edit mode.

To filter the list of all guardians by a tag, type the tag in the Search by name or tag field—the page automatically updates to show only guardians with matching tags.

This DQL shows you the first guardian.validation.objective business event with a specific guardian ID and parses the guardian tags field to extract a specific tag value from the event JSON.

fetch bizevents |
filter event.type == "guardian.validation.objective" AND guardian.id == "vu9U3hXa3q0AAAABADFhcHA6ZHluYXRyYWNlLnNpdGUucmVsaWFiaWxpdHkuZ3VhcmRpYW46Z3VhcmRpYW5zAAZ0ZW5hbnQABnRlbmFudAAkMWNiZDVkYWYtZThhNi0zMDkxLWFkOGQtMmU5NDNmNWJmZWJmvu9U3hXa3q0" |
limit 1 |
parse guardian.tags, "JSON:parsed_guardian_tags"

This DQL shows you all guardian.validation.finished business events from guardians tagged as tagkey:my-tagged-guardian.

fetch bizevents
| filter event.type == "guardian.validation.finished"
| expand guardian.tags
| filter contains(guardian.tags, "my-tagged-guardian")

You can automate the execution of a guardian via Workflows, tying guardian execution to an event.

To add a guardian action to an existing workflow

You can create a new workflow by selecting Automate on the top right of the guardian page. When you create a workflow this way, the following parameters are configured, but be sure to adapt them as needed.

The guardian action generates the following output and passes it to the subsequent actions of the workflow.

To learn more about workflows for a guardian, open the help menu in the upper-right corner of a guardian and select Get started with Automation.

If a workflow is created, your guardian is validated automatically. You can also perform the validation manually.

Validation overview

By default, the overview page shows validations for the last seven days. You can view older results by opening a guardian and selecting a different timeframe.

Automated validation

The event subscriptions in the workflow define when the validation of a guardian has triggered automatically.

Manual validation

You can perform a validation of a guardian by selecting the Validate button on the overview screen or within the validation details screen.

Individual objective result

For each objective, the validation returns the derived value and classification. The severity goes from the highest (1) to the lowest (5).

Overall validation result

After the validation of each objective is done, the guarding uses the most severe of individual validations as the overall validation result. Examples of this result usage include: