Configure assertions

An assertion is a team-scoped evaluation linked to a monitor. Each time a check log arrives, linked assertions run and record an assertion run as healthy or failure. Failure includes a clear error message.

Use assertions to decide when a probe result should open or close an incident. Do not use this page for Check Script (DSL); that is a separate Starter and Standard feature on the monitor itself.

Before you begin

• Create a monitor.
• Plan caps apply: assertions per team (10 / 20 / 50) and assertions linked per monitor (1 / 3 / 5). See plan limits.

Link an assertion

When you create a monitor, the wizard includes an Assertions step after Active monitoring. You can also link an assertion later from the monitor.

1. Open the monitor and choose Link assertion, or create an assertion in the team’s assertion library first. On create, complete the Assertions step in the New Monitor wizard.

2. Set hysteresis, or keep the defaults:

   • failure_count = 3
   • window_size = 5
   • recovery_streak = 5

3. Save. Changes to the monitor apply to every linked monitor–assertion pair.

Hysteresis parameters must satisfy 1 <= failure_count <= recovery_streak <= window_size <= 10. Opening an incident is easier than resolving it: recovery requires a consecutive success streak, not merely fewer failures in the window.

How incidents open and close

For each monitor–assertion pair:

• An incident opens when the assertion run state swaps from healthy to failure, after the failure pattern meets hysteresis.
• That incident closes when the state swaps from failure to healthy, after the recovery streak.
• At most one incident is open at a time for that pair. If an incident is already open, Crystade does not open another.
• If an incident is already open and the latest response is investigating, a new incident response is inserted only when the rendered content is different.

Verify it worked

• Trigger a failing check (or wait for one). After failure_count failures in window_size checks, an incident appears.
• Restore the target. After recovery_streak consecutive successes, the incident resolves.
• Confirm only one open incident exists for that monitor–assertion pair.

Note

The first-win onboarding path requires at least one linked assertion so you can see assertion-run pass/fail next to the probe outcome.

Troubleshooting

Symptom	What to check
Cannot link another assertion	You hit the per-monitor or per-team cap for your plan.
Incident opened too quickly	Raise failure_count or window_size.
Incident never resolves	Recovery requires recovery_streak consecutive successes, which must be greater than or equal to failure_count.

Related

How monitoring relates to incidents Why hysteresis exists and when to use a manual incident.

Report and update incidents Severity, responses, and publish.

Create an availability monitor Protocols, intervals, and check logs.