Alert Config LabLearn how alert conditions work · Beginner exercise with answers
Rolling window vs. retest window
The rolling window selects the time span processed by the aligner. The retest window specifies how long processed values in the same series must continuously violate the condition. A wider processing window is not the same as waiting longer before an alert opens.
Make a prediction
The first value above 5% appears at 60 seconds in the table below. Which retest setting opens an alert: 60 or 120 seconds? Predict the opening time too.
A fixed, small input example
Compare retest-window behavior on one fixed waveform with a brief violation and recovery.
| Elapsed time | Observed 5xx ratio |
|---|---|
| 0 s | 0% |
| 60 s | 6% |
| 120 s | 6% |
| 180 s | 0% (recovered) |
Fixed public teaching seed: 270927 · The dataset is fixed by the table above; it contains no personal settings.
How values flow into the evaluation
- Observed value at each timestamp
- ALIGN_NONE: pass each value through
- Check whether one series stays in violation for 60 / 120 seconds
Conditions held fixed in this example
- Threshold > 5%; ALIGN_NONE; one series. Compare 60-second and 120-second retest windows; all other settings stay fixed.
- This synthetic fixture is independent of the app's current settings.
fixed-teaching-points · v1 · 60 seconds sample interval · 180 seconds horizon
Answer and explanation
Reveal the answer
In this evaluator, the 60-second retest opens at 120 seconds; the 120-second retest does not open before recovery.
The violation starts at 60 seconds. By 120 seconds, it has lasted 60 seconds, satisfying the shorter retest. At 180 seconds the value is already 0%, so the 120-second continuous-violation requirement is not met. Check evaluation timestamps and continuity, not just the number of violating points. This fixture uses ALIGN_NONE to isolate retest behavior from rolling-window effects.
Explore in the simulator
This is a separate experiment using your current settings, not an import of the fixed table. Review your rule first and, if needed, build a practice waveform before trying these steps.
- Review the current rolling window, aligner, and retest window. Note the dataset ID.
- Change only the retest window, for example from two minutes to one. Apply it and compare with the previous rule.
- Check alert opening times, cases where no new alert opened, and retest resets. Confirm that the dataset ID stayed the same.
Opening the app shows a confirmation dialog. The link alone does not change settings, waves, or saves. If you start a teaching test, the original experiment is protected in a separate browser-storage area. The fixed table above and the simulator's test for your current rule are separate datasets.
Check your understanding
- Explain the different roles of the rolling window and retest window.
- Explain why the duration of a raw spike alone cannot determine whether an alert opens.
- Check that the dataset ID stays the same when comparing rules; do not regenerate the waveform.
For a comparison note, record the dataset ID, changed setting, and opening time or reason no alert opened. Save the experiment explicitly in the app if you want to return to it.
Scope and limits
This isolates retest with no rolling window. For a separate one-minute window example, 1%, 1%, 1%, 9% gives MEAN 3% and MAX 9%. Raw spike duration alone does not determine an alert.
Educational comparisons use synthetic data. They do not guarantee production detection rates or notification timing, and do not connect to or configure Google Cloud. Unsupported operations and insufficient history can be unevaluable.
Official references
Reviewed September 27, 2026. Not an official or affiliated Google service.
Return to the simulator
← Back to the simulatorIf you opened this guide in a new tab from the app, close this tab to continue in the original tab. The link above opens the simulator again. To resume the same waveform and playback position later, use Save / open experiment in the app. This is separate from settings autosave.