Skip to content

Get alerted on a problem

Nobody can watch a dashboard all shift. An alert rule watches it for you and speaks up when something crosses a line you set.

  • Data already flowing — detection set up or devices reporting telemetry
  • A rough sense of what “normal” looks like for the thing you are watching

Open Explore and chart the metric over the last week.

This step gets skipped and it is the reason most alerts end up muted. A threshold chosen without looking is either so tight it fires constantly or so loose it fires after the damage is done. Two minutes here saves a month of noise.

Open Alerts and select Create Rule.

The Create Alert Rule dialog with name, SQL condition, operator, threshold, severity and evaluation interval

Fill in:

Field What to put
Rule Name What an operator would need to read at 3am — “Defect rate above 4% on Line 3”
Description Optional. Why this matters and what to do about it
Condition SQL The query that produces the number to test
Operator How to compare it — greater than, less than
Threshold The number from step 1
Severity Info, Warning or Critical
Evaluation Interval How often to check, in seconds. Defaults to 300 (five minutes)
Severity Means
Critical Stop the line. Wake someone up
Warning Deal with it this shift
Info Worth recording, not worth interrupting anyone

Severity is a promise to whoever receives it. If everything is critical then nothing is, and people stop reading.

Rules are evaluated in the background — nobody needs a browser open.

Watch it for a shift or two. If it fires when nothing was wrong, raise the threshold or lengthen the interval. Rules can be disabled without deleting them, which is what you want during planned maintenance rather than deleting and rebuilding.

The alert appears under Active Alerts. Acknowledging it marks it as seen.

If it needs actual work, raise an incident — an alert says something is wrong, an incident tracks what you did about it.