Alert studio

Draft · rule-0142 · never published

Definition

edits apply live

Rule name

What to do about it This shows up inside the alert when you open it at 2am.

Labels Comma separated. Used for filtering and grouping.

Runbook link

Trigger

Metric

Condition

Threshold Try 8 through 34 and watch the match count.

Check everyHow often Proxima samples the metric.

Must hold forHow long the condition stays true before it fires.

Group alerts by One alert per group instead of one per match.

Compare against

Close it automaticallyResolve the alert when the metric comes back to normal.

Scope

Watch

Except Names or patterns to leave out. CI runners are noisy by design.

Stay quiet during maintenanceSkip anything inside a scheduled window.

Stay quiet while a host rebootsGive a rebooting host two minutes before complaining.

Delivery

SeveritySets the pill, the channels and whether it can wake you.

ChannelsCritical always adds a phone push, whatever is set here.

Push Email Webhook

Remind me again after

Outside waking hours

Start it pausedSave the rule but do not let it fire yet.

Preview

last 24 hours

How the alert will look

High 12 services now

Slow responses on argon

Response time p95 (ms) goes above 24 for 5 m

What to do. Check caddy first, then the database connection pool on kryptonane.

pushemail

Backtest

The metric over the last 24 hours. The dashed line is your threshold.

Would have fired

12×

Verdict

Sensible

What it is watching

A sample of the current selection

caddy

argon 14 ms p95, well under

api

argon 27 ms p95, over the line

web

argon 19 ms p95, well under

argon 31 ms p95, worst offender