ERROR-BUDGETS-DETAIL

Failure budget = 1 - SLO.

Definition

If SLO = 99.9% over 30d, error budget = 43min 12s. If consumed mid-month, freeze releases until reset. Otherwise, authorisation for risky deploys. Implemented via burn-rate alerts (accelerated consumption).

Origin

Google SRE concept, formalised in Site Reliability Engineering 2016 chap. 3.

Example in context

An EDI service consumes 80% of its error budget in 2 weeks — release freeze until month end.

Related terms

• Burn rate — related alert.