Observability & Evaluation · Software component
Error Budget Tracker
Software componentObservability & EvaluationObservability & Evaluationarc:ErrorBudgetTracker
An SLO-management component that accumulates failures against the error budget implied by an SLO over a trailing window and computes the burn rate at which that budget is being consumed.
Responsibility. Computes remaining error budget and current burn rate for each SLO window.
Also known as: Error budget calculator, Burn rate monitor
Relationships
is configured by structural
reads dependency
sends data to dynamic
- Alert Manager Ch8.2A
- Metrics Dashboard abstract Ch8.2A
Design guidance
- SHOULD derive alert urgency from burn rate (current error rate divided by the SLO error threshold) rather than from absolute error rate.
- SHOULD accumulate the error budget over the full SLO window rather than enforcing hourly limits.
- SHOULD require the burn rate to be sustained over a time window, longer for lower severities, before alerting to avoid paging on brief spikes.
- SHOULD keep an SLO in breach until high-error periods roll out of the trailing window.
Quantitative guidance
As stated by the sources; verify before use.
- Burn rate = current error rate / SLO error threshold; 1.0 exhausts the budget exactly at window end, 2.0 in 15 days, 5.0 in 6 days, 10.0 in 3 days, 20.0 in 1.5 days (Ch8.2A).
- 99.9% availability over a 30-day window = 0.1% error budget = 43.2 minutes of 43,200 minutes (equivalently 3.6 min/hour, not enforced hourly) (Ch8.2A).
- Scenario: 0.05% error rate (burn 0.5) on Days 1-5 consumed ~3.6 min; 0.5% (burn 5.0) on Days 6-10 consumed 36 min at 7.2 min/day; budget exhausted midday Day 11; SLO restored ~Day 35 as bad days left the window (Ch8.2A).
Classification
- Patterns
- Error budgetBurn-rate alertingMulti-window burn-rate alertingTrailing-window SLO
- Technologies
- Prometheus
- Quality attributes
- Reliability (ISO/IEC 25010 | NIST AI RMF: valid and reliable)
- Risks mitigated
- Silent error-budget exhaustionLate alerting on absolute error-rate thresholdsAlert fatigue from minor fluctuations
Sources
- Ch8.2A: T. Nguyen, "Error Rates and Reliability," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 8.2A. ISBN: 9798244538229.