Reliability
Platform status, monitoring & backups
A live health probe plus the monitoring, alerting and recovery commitments behind it — the page a lender's IT reviewer asks for before signing.
Live probe refreshes every 60 seconds
Uptime target
99.9%
monthly
Data loss ceiling
5 min
RPO
Restore target
4 hrs
RTO, full platform
Probe frequency
60 sec
multi-region
Monitoring & alerting
What we watch, and what wakes someone up
GET /api/public/health every 60s from multiple regions
2 consecutive failures → page on-call
Connection, latency and error-rate sampling
p95 query latency > 800ms for 5 min → alert
Nightly golden-path re-verification of every sealed credential
Any failed verification → tenant admin notified
Sign-in success rate and session errors
Success rate < 95% over 10 min → page
Payment event delivery and processing failures
Any unprocessed webhook > 15 min → alert
Client and server exception capture with release tagging
New error affecting > 5 sessions → alert
Backups
Encrypted, tenant-isolated, restore-tested
Point-in-time recovery to any moment in the retention window
Encrypted at rest, retained 7 days on the managed platform
Retained 90 days for regulatory restore requests
Object storage with versioning and tenant-scoped path isolation
Disaster recovery
Objectives, testing and tenant-level rollback
5 minutes — the maximum data loss in a total-failure scenario, backed by continuous write-ahead-log archiving.
4 hours for full platform restoration; 1 hour for a single-tenant restore from snapshot.
A restore drill is performed quarterly against a scratch environment, and the result is recorded in the audit ledger.
A single workspace can be rolled back to a point in time without affecting other tenants, on request through Sev-2 support.
Compute runs on a global edge network with automatic failover; the database platform supports cross-zone recovery from snapshot.
For your monitoring team
Point any external uptime service here
GET https://nexusmortgageos.com/api/public/health → 200 { "status": "ok" } · 503 when degradedThe endpoint is unauthenticated, returns no customer data, and is safe to poll every 30–60 seconds. Incidents are posted here and emailed to workspace administrators per the Support & SLA policy.