Alert rules that come with a diagnosis
A threshold tells you something crossed a line. Varqeni then looks at the system the rule belongs to, the way it does for anything it finds itself, so the page says why.
AvailabilityAvailable to everyone with early access to Varqeni.
How it works
- Write the rule: pick a system, then either fill in a metric, above or below and a number, or write PromQL that returns a value only while the condition holds. Set how long it must hold before it fires.
- Checked before it is saved: Varqeni asks Prometheus first. A rule that reads more than a day of data on every check, matches every series, or returns too many series is refused with the reason.
- Try it on the past: a backtest replays the rule over the last days and lists the moments it would have fired, beside the incidents Varqeni already opened in that time.
- When it fires: a page rule opens an incident on its system. Varqeni collects and seals the evidence and names the cause when its checks see one; when they see nothing, the rule and its value are the incident. A notice rule is told in chat and never wakes anyone.
- When it clears: one check without a match resolves the incident and its page.
- Quiet on purpose: silences and maintenance windows hold rule pages like any other.
- Who may change them: everyone signed in sees the rules; an admin on the rota writes, enables, disables and deletes them. Every change is kept in a sealed history.
What you should see: The rule, its expression and the value that fired it are part of the incident's sealed record. In the Fortis lab a rule on Payments API having no ready replica fired when the service was scaled to nothing, opened an incident marked as coming from the rule, and Varqeni's own investigation named the cause (the endpoint does not answer); the backtest found the moment and the rule's history checked out against its locked copy.
What it can and cannot do
Can
- Watch any metric you choose, investigate when it crosses the line, and page with the cause and evidence; check a rule against the last days first.
Cannot
- Save a rule that would read too much on every check, or wake anyone for a notice rule.
Related articles
- RulesEvery alert rule on one page: what it watches, whether it is firing, its value over the last day, and when it would have fired before.
- Paging and escalationVarqeni pages the person on call only when it cannot settle an incident itself, and escalates when nobody answers.
- Alerts from Prometheus Alertmanager and other monitoringSend the alerts you already have to Varqeni. Each one opens an incident on the system it is about, with that system's evidence sealed beside it.
- GrafanaKeep Grafana for your dashboards. Varqeni marks its incidents, deploys and fixes on your charts and links every incident to the right panel.
Still stuck? Contact support or email info@fortisdatagroup.com.