Zimin A.N.
bcedddef71
docs(ops): SLO/SLI definitions + error budget framework
7 SLI с конкретными порогами и justification:
- read-api availability (99.5% v1 target)
- read-api p99 latency < 500ms target / 1s alert
- writer 5xx error rate < 1%
- outbox publish lag (≤ 50 normal, alert > 100)
- outbox DLQ size = 0 strict
- outbox publish error rate ~0
- postgres availability (99.9% v1 target)
Error budget framework для 99.5% уровня (~216 мин/мес downtime).
Burn rate alerts — TODO.
Notification routing — TODO (Slack/Telegram per severity).
Что не покрывает v1: Grafana dashboards, distributed tracing, log
aggregation, capacity planning. См. observability stack roadmap для v2.