This lesson's interactive features are locked, please to keep using them

Alert Thresholds

Setting sane thresholds is what makes alerting practical.

A rule should fire only when a bad condition is sustained long enough to matter. In Grafana, that usually means tuning:

The threshold value (how many 500-status responses is enough to trigger an alert)
The query window (over what time period we need to see the threshold breached)
The "pending period" duration (how long the condition must hold)

We want to avoid firing when a short spike appears and suddenly disappears (for example, a blip in a deployment that self-heals).

Assignment

Update your alert rule's thresholds and pending duration.

Open the alert rule for editing and change the "pending period" in the evaluation behavior section to 20 seconds and save the rule.
If necessary, wait for up to 5 minutes for the rule to be in a "normal" state.
Trigger a brief burst of 25 500 errors by spamming login attempts with "saruman"/"malformedPassword", and refresh the alert rule page.
Verify the rule is in a "pending" state
Wait for 20 seconds, refresh the page, and verify the rule is now in a "firing" state.

Run and submit the CLI tests from the root of the Linko repo.