Latency alerting
Configure latency detection, learn the alert states, and avoid false incidents.
Latency detection evaluates slow responses separately from availability failures. Configure it when response speed is an operational requirement.
Choose a detection mode
| Mode | Behaviour |
|---|---|
| Off | Stazus measures latency but does not classify anomalies or open latency incidents. |
| Shadow | Stazus classifies anomalies for review but does not open incidents. |
| Automatic | Stazus learns a baseline from recent checks and evaluates anomalies against it. |
| Fixed | Stazus evaluates latency against the fixed threshold that you set. |
Start with shadow mode
- Create or edit an endpoint.
- Select Shadow as the latency detection mode.
- Save the endpoint.
- Review the endpoint Alerting panel after representative checks run.
- Select Automatic or Fixed when the results match your operational expectations.
Shadow mode reduces the risk of alert noise while you learn normal performance.
Configure automatic mode
Automatic mode learns a baseline. The Alerting panel shows its sample count, median, p95, open threshold, and recovery threshold.
Set these values:
| Setting | Limit or meaning |
|---|---|
| Sensitivity | A number from 1 to 20. |
| Minimum learning samples | An integer from 20 to 10,000. |
| Trigger breaches | The number of breaches needed to trigger. It cannot exceed the trigger window. |
| Trigger window | The number of checks examined for breaches. It is 1 to 20. |
| Recovery streak | The number of healthy checks needed for recovery. It is 1 to 20. |
| Fixed threshold | An optional safety ceiling during automatic mode. |
Stazus shows Warming up until it has the required number of samples. If no learned baseline exists and you set a fixed threshold, Stazus uses that fixed ceiling.
Configure fixed mode
- Edit the endpoint.
- Select Fixed threshold.
- Enter a threshold in milliseconds.
- Set the trigger and recovery values.
- Save the endpoint.
The fixed threshold is required in fixed mode. It must be at least 1 millisecond.
Read alert states
The Alerting panel can show these signal states:
| State | Meaning |
|---|---|
| Normal | Stazus has no active signal. |
| Suspected | Stazus observed a recent breach but did not open an incident. |
| Firing | Stazus has an active incident for the signal. |
| Recovering | The active signal has a recovery streak. |
Use the panel with check history. A latency anomaly does not always prove that users experienced an outage.
Related guides
Last updated on