Certificate Lifecycle Service · View 16 of 21 · Operations
Decisions
- Probers are deployed per region, outside the primary control plane's failure domain, so a control-plane outage does not also blind verification.
- CRL and OCSP are published globally through CloudFront and held to 99.99%, because relying parties the platform does not control depend on them.
- The root lives in a ceremony facility on no network. Its recovery path has a human in it by design: RTO ≤ 72 hours.
Fail static
- A control-plane outage stops issuance and stops nothing else. Existing certificates keep serving; orders queue.
- Estate survival without the control plane: ≥ 8 hours, which is the renewal margin on a 24-hour leaf. The platform publishes this number rather than discovering it during an incident.
Assumptions
- Primary eu-west-1 across 3 AZ; eu-central-1 and us-east-1 carry a warm standby control plane and a prober.
- Registry RPO 0 / RTO ≤ 30 min; issuing intermediate replacement live and distributed ≤ 4 hours, which is why two are kept live.