9 Things You Should Do When Starting to Use SLOs
summary
Nine concrete SLO practices: pick measure meaningful things tied to user journeys, export raw data (no preaggregation), prefer success rates over error rates, and back objectives with history — e.g. a payment acknowledgment SLA set at 99% within 2000 ms with observed P50 ≈100 ms, P90 ≈1000 ms, P95 ≈1500 ms. Catch hidden patterns like a 24‑hour cache refresh causing 30–40 minute latency spikes, tie SLOs to runbooks and automation, document definitions, and adopt OpenTelemetry/OpenSLO.
AI-written summary · about
chapters (13)
- 0:00SLOs introduction
- 1:16SLOs implementation challenges
- 3:01Reliability culture with SLOs
- 4:05Systems via meaningful metrics
- 8:19Success vs error rates
- 10:03Raw data for SLIs
- 13:17Setting objectives based on historical data
- 17:59SLOs and automated actions
- 20:12SLOs for different personas
- 23:59When systems fail SLOs
- 27:30Documenting SLOs with context
- 28:53Using SLODLC for repeatable SLO development
- 29:59SLO communities
related talks
28mLean Product Development Through SLOs
35mAvoiding Goodhart's Law - Use SLO's as Tools Not Cudgels
06ma pint sized introduction to SLO
38mImplementing SRE practices: SLI/SLO deep dive
41mSLOs that Delight: Master Multi-Service SLO Calculations
24mSLO Creation And You: Or, How We Learned To Stop...
22mLook Ma, No Metrics: A Quest Towards Extracting SLOs
33mHow and why we lowered our SLO