Track and create SLOs from your service entities
Track and create SLOs from your service entities
Every Service entity in the entity catalog carries an SLOs tab, so you can check whether a service is meeting its service level objectives (SLOs) while you investigate it, without switching to another app. The entity header carries an SLOs dropdown menu with a status dot, which gives you the same answer at a glance.
Each objective appears as a card with its measured service level indicator against target, the share of error budget remaining, and the absolute budget left in the compliance window. A Burning badge, and a banner at the top of the tab, tell you when a service consumes its error budget faster than allowed.
Let the assistant propose objectives from real traffic
For a service with no objectives, select Generate SLO proposals. Grafana Assistant reads the service’s recent request, error, and latency metrics and proposes availability and latency objectives grounded in how that service behaves. Review each proposal and adjust Target (%), Window (days), Threshold (ms), and an optional Team before anything is created. On a service that already has objectives, select Add SLO and the proposals complement what exists instead of duplicating it.
Proposals become real Grafana SLOs, labeled with the service they belong to, so they evaluate and appear in Grafana SLO alongside every other objective. Each one is created with Grafana SLO’s default burn-rate alert rules. You can also select Create manually to define an objective yourself.
Move between the entity and the objective
From any card, select View performance to open that objective’s performance dashboard, or Edit SLO to change its definition. Values read Pending until the backend evaluates the objective for the first time.
Requirements
This capability applies to Service entities in Grafana Cloud with the Knowledge Graph active, and is rolling out to Grafana Cloud stacks in stages, so it might not have reached your stack yet. AI-generated proposals also need Grafana Assistant on the instance, and a service with request, error, and latency metrics. Where either is missing, you can still define an objective manually.
