SLI → SLO → Error Budget

Nine nines don't exist. A good SLO, does.

Wrong window. Wrong burn rate. A forgotten business-hours filter that quietly breaks your alerting. The wizard bakes Google SRE defaults in from the start, so you skip all of that.

Start the wizard →

Not a mockup — the actual rule

Real output from the wizard, footnoted below. Checked by promtool check rules before it reaches your ZIP.

recording-rules.yaml — api, 99.90% availability# window: 30d rolling1 record: business_hours:active2 expr: (hour() >= 8 < 18) * (day_of_week() >= 1 <= 5) record: sli:api:availability:ratio5m:bh3 expr: sli:api:availability:ratio5m and on() business_hours:active == 1

The right window. 30 days, not a noisy 5-minute ratio that pages someone over an hour of bad luck.

The hour() trap, solved. Prometheus evaluates hour() at query time — this rule freezes it into a real recording, not a live footgun.

The forgotten filter. Office-hours gating that's usually hand-written once, then quietly rots.

excerpt — 4 rules generated, 2 shown ✓ promtool check rules — passed

How it works

Define

Pick your service and SLI kind. Five quick-start presets cover REST APIs, gRPC, Kafka, pipelines, and blackbox probes.

Set your target

Choose an objective and window. Google SRE burn-rate defaults are pre-filled — edit them if your team runs differently.

Download

A validated ZIP: recording rules, alerts, Grafana dashboard, runbooks — promtool-checked before it reaches you.

Why nines

Google SRE-compliant

14.4×, 6×, 1× burn rates — the actual numbers from the book, not adapted.

promtool-validated

Every recording rule and alert is checked before it reaches your ZIP.

Business-hours aware

Filter SLIs to office hours without hand-writing the hour() trap yourself.

Start the wizard →

Start the wizard

Pascal

SRE and observability on the banking side, twenty years of telecom BSS/OSS before that.

you.arewel.com