An error budget is the allowable unreliability remaining under a service level objective (SLO) during its measurement period. It shows how much wiggle room remains before the agreed goal is breached; burn rate measures how quickly that allowance is being consumed. section 10.0 section 10.1.3
A monthly household target
Suppose a child must keep dirty clothes on the floor to three or fewer pieces during a month. Each inspection updates the measurement, while the remaining allowance shows how close the goal is to being breached. The same pattern applies to service reliability: measurements are compared with an SLO across a defined period. section 10.0
Error budgets turn reliability goals into an operational signal. Teams can use them to improve alert accuracy, reduce noisy or false-positive alerts, and support trend-based warnings linked to runbooks or automated remediation. The model is useful but optional; many organizations track SLIs against SLOs without adopting error budgets or burn rate. section 10.1.3 section 10.4
Listing 10.3 Pyrra SLO definition
apiVersion: pyrra.dev/v1alpha1
kind: ServiceLevelObjective
metadata:
name: ho11y-api-query #1
namespace: monitoring
labels:
prometheus: k8s
role: alert-rules
spec:
target: '75.0' #2
window: 7d
indicator:
ratio:
errors: #3
metric: ho11y_total{job="svc",http_status_code=~"4.."}
total: #4
metric: ho11y_total{job="svc"}
A Pyrra ServiceLevelObjective definition makes the budget computable by declaring a target, a measurement window, and error-versus-total Prometheus metrics. section 10.2.2


