pal-e-docs

Phase 16: SLO Governance (Sloth)

phase-pal-e-platform-16-slo-error-budgets Phase

Project

Phase 16: SLO Governance (Sloth)

Goal: Deploy Sloth for Prometheus-native SLO generation, define SLOs for all critical services, and create error budget burn rate alerting — the measurement layer that gates every subsequent tier.
Owner: Dev agent
Repo: forgejo_admin/pal-e-platform
Depends on: Phase 14 (synthetic monitoring provides availability data), Phase 15 (DORA re-baseline establishes metric targets)
Tier: 1 — Foundation. SLOs gate everything. No load test, chaos experiment, or canary promotion is meaningful without SLO baselines to evaluate against.
DORA: All four metrics. SLOs formalize CFR and MTTR targets. Without SLOs, "Elite" is subjective. With SLOs, it's a number.

Scope

  • Deploy Sloth via Helm — generates multi-window multi-burn-rate Prometheus recording rules from human-readable YAML (the Google SRE book pattern, automated)
  • Define Critical User Journeys (CUJs) as the basis for SLIs:
  • Define SLOs per service tier:
  • Grafana SLO dashboard: error budget burn rate, remaining budget, compliance history per CUJ
  • Alert when error budget is <20% remaining (slow burn ticket) or <5% (fast burn page)
  • SOP: sop-slo-governance — how to define, review, and update SLOs

Deliverables

  • TBD — filled after completion
  • plan-pal-e-platform — parent plan
  • phase-pal-e-platform-14-synthetic-monitoring — provides the probe_success data SLIs consume
  • phase-pal-e-platform-15-dora-rebaseline — provides DORA metric baselines
  • phase-platform-21-progressive-delivery — Argo Rollouts gates on SLO burn rate
  • phase-platform-22-load-testing — load tests validate capacity against SLO targets
  • phase-platform-23-chaos-engineering — chaos experiments prove SLOs hold under failure