Skip to content
Observability Stack

SLOs

Service Level Objectives (SLOs) turn a reliability target — “99.9% of GenAI invocations succeed over 28 days” — into deployed Prometheus recording rules, multi-window burn-rate alerts, and a live error-budget view. The SLO app builds all of that from a guided wizard; you never hand-write PromQL unless you want to.

The SLO catalog with its health overview, then opening a breached SLO to reveal attainment below target, an exhausted error budget, and the four burn-rate tiers

The SLO catalog and health overview, then into a breached SLO — attainment below target, budget exhausted, and the multi-window burn-rate tiers.

Concepts

Section titled “Concepts”
TermMeaning
SLI (Service Level Indicator)The measured ratio of good events to total events — for example, non-error requests ÷ all requests.
SLO (Service Level Objective)The target the SLI must meet over a window, e.g. 99.9% over 28d.
Error budgetThe allowed shortfall: 100% − target. A 99.9% SLO permits 0.1% bad events.
Burn rateHow fast the budget is being consumed relative to sustainable. 1x exhausts the budget exactly at the window’s end; 14.4x exhausts a 28-day budget in ~2 days.
MWMBRMulti-Window Multi-Burn-Rate alerting — the Google SRE pattern of pairing a short and a long window per tier to catch fast burns quickly and slow burns reliably without flapping.

How it works

Section titled “How it works”

Each SLO you create is compiled into a Prometheus rule group deployed to the Prometheus rule engine:

  • Recording rules — pre-compute the SLI error ratio at multiple time windows (5m through 3d).
  • Burn-rate alerts — four severity tiers that fire only when both a short and a long window exceed the tier’s burn threshold (the MWMBR pattern).
  • Error-budget warnings — fire at configurable budget thresholds (e.g. 50% and 20% remaining).

The SLO app reads those same rules back to power the error-budget, burn-rate, and request-volume charts, and the burn-rate alerts flow into the Unified Alerts View alongside your other alerts.

Finding your way around

Section titled “Finding your way around”

SLOs is in the side navigation under Application Performance.

  • SLO catalog (landing page) — every SLO with owner, objectives, rule status, and health, plus a health-overview strip and a deep filter rail (state, SLI type, canonical kind, service, team, tier, mode, enabled).
  • Create an SLO — the template-first wizard.
  • Explore an SLO — the detail page: attainment, error budget, burn-rate tiers, and the generated rules.
Section titled “Related”