Reliability technical appendix
This page is for technical decision makers (IT, security, procurement) who want the architecture behind the reliability overview. It documents where RentalTide runs, what we replicate, how fast we recover, and what sits outside our direct control. It describes what is in place today, including the gaps.
Hosting and regions
RentalTide is hosted on Amazon Web Services. We do not operate any on-premise servers.
| Component | Region | Notes |
|---|---|---|
| Application servers (API) | us-east-1 | Containers on ECS Fargate across two Availability Zones, autoscaled from 2 to 20 |
| Primary database (Aurora PostgreSQL) | us-east-1 | One database instance; storage replicated six ways across three Availability Zones |
| Object storage (photos, attachments, PDFs) | us-east-1 | S3; versioning on the buckets holding signed waivers, onboarding documents and database archives |
| Identity images for Australian locations | ap-southeast-2 (Sydney) | Stored in Australia; every other location's images are stored in us-east-1 |
| Background jobs and scheduled tasks | us-east-1 | Run inside the application service; each job holds a database lock so only one server runs it |
| Web app, booking widget and docs | Global edge | Amazon CloudFront |
| API access from Australia and New Zealand | Edge in the region | CloudFront relays requests to us-east-1 and caches only public, non-personal responses |
| Backups | us-east-1 | Same region as the data they protect, across multiple Availability Zones. No copy in a second region |
The web app and booking widget are static files served from CloudFront's edge, so they keep loading during a us-east-1 event. Anything that needs the API (availability, checkout, the admin panel's data) does not work until the API is back.
Database resilience
The primary database is Aurora PostgreSQL on a single provisioned instance.
- Storage across three Availability Zones. Aurora writes every change to six copies across three zones and acknowledges a write once a quorum has it. Losing one zone's storage loses no data.
- One database instance, no standby reader. If the instance fails, Aurora recreates it automatically, usually in under 15 minutes. The service is unavailable while that happens. A standby reader in a second zone is on our roadmap and is not in place today.
- Continuous backup. Point-in-time restore is available to any second within the last 35 days.
- Encryption. The database and its backups are encrypted with a customer-managed AWS KMS key, and deletion protection is on.
- No cross-region copy. Backups stay in
us-east-1.
Application tier
The API runs as containers on AWS ECS Fargate behind an Application Load Balancer.
- Two to twenty containers across two Availability Zones. Losing one zone removes part of the capacity; autoscaling replaces it in the other zone.
- Health checks every 10 seconds. A container that fails three checks in a row is taken out of rotation and replaced.
- Zero-downtime deploys. New containers must pass health checks before they take traffic, and the old ones keep serving until they do.
- Automatic rollback. If a new version keeps failing its health checks, the deploy rolls back to the previous version without manual intervention.
Recovery time and recovery point targets
Recovery time objective (RTO) is the target time from the start of an incident to service restoration. Recovery point objective (RPO) is the target for data lost, measured in time. These are engineering targets, not guarantees; our contractual commitments are in the Service Level Agreement.
| Failure scenario | RTO target | RPO target | Mechanism |
|---|---|---|---|
| Single application container crash | About 1 minute | Zero | Health check removes it; the other containers keep serving |
| Failed deploy | Minutes | Zero | Old containers keep serving; automatic rollback |
| Availability Zone failure (application servers) | Minutes | Zero | Containers in the other zone; autoscaling refills capacity |
| Database instance failure | Usually under 15 minutes | Zero | Aurora recreates the instance on durable storage |
| Application bug that corrupts data | Hours, depending on detection | Bounded by detection time | Targeted repair from the journal_entries ledger |
| Catastrophic data corruption | Several hours | About 5 minutes | Point-in-time restore from continuous backup |
Full region outage (us-east-1) | Until AWS restores the region | Depends on the region's storage | No standby region today; backups are in the same region |
We have not yet run a formal disaster recovery exercise. Until we do, the database targets above are based on AWS's documented behaviour rather than our own drills.
Audit and reconstruction
Beyond database backups, RentalTide keeps an append-only, double-entry journal (journal_entries) of every financial change: payments, refunds, fees, taxes and account adjustments.
This means:
- The amount paid, outstanding balance and amount refunded on a booking can be rebuilt from the journal at any time.
- A bug that corrupts those cached figures without touching the journal is fixed by re-running the sync from the journal.
- Forensic questions (who paid what, when, on which booking) survive any cache corruption.
External dependencies
RentalTide depends on a handful of external providers. When one has an outage, the feature it powers degrades; the rest of RentalTide keeps working.
| Provider | What we use it for | Failure behaviour |
|---|---|---|
| Stripe | Payments, terminals, Connect | Card capture and refunds fail with clear errors; no data loss |
| SendGrid | Transactional email | Email is delayed; SendGrid retries deferred mail |
| Twilio | SMS and voice | Texts and calls fail or are delayed |
| AWS | Compute, storage, database | See the scenarios above |
| OpenAI / Anthropic | AI features (docs bot, phone, summaries) | AI features stop; core operations are unaffected |
Staff and merchant sign-in is run by RentalTide on its own AWS infrastructure, not by a third-party identity provider.
Backups and retention
| Asset | Frequency | Retention | Storage location |
|---|---|---|---|
| Aurora continuous backup | Continuous | 35 days | AWS-managed, us-east-1 |
| DynamoDB point-in-time restore | Continuous | 35 days | AWS-managed, us-east-1 |
| S3 object versioning | Every write | 90 days per version | S3, us-east-1 |
| Application logs | Real time | 400 days | CloudWatch, us-east-1 |
| Financial journal | Real time | Indefinite | Aurora journal_entries table |
S3 versioning covers the buckets holding signed waivers, onboarding documents and database archives; a deleted or overwritten file in those buckets can be recovered for 90 days. Cross-region and cross-account backup copies are not in place today.
Security controls relevant to availability
- DDoS: AWS Shield Standard is always on, CloudFront absorbs volumetric traffic at the edge, and AWS WAF filters requests in front of the API.
- Credential theft: sign-in tokens are short-lived, multi-factor authentication is available on every account, and repeated failed sign-ins lock the account.
- Monitoring: AWS CloudTrail records account activity, Amazon GuardDuty watches for threats, and security alarms page our on-call engineer around the clock.
- Staff access: when RentalTide staff open a merchant's account for support, the session is time-limited and logged.
RentalTide itself does not hold SOC 2, ISO 27001 or PCI DSS certification. The providers above do: see Security and compliance.
Status, communication, and runbooks
When something goes wrong, you will hear from us here:
status.rentaltide.com: the public status page and the source of truth for the platform overall.- Your support channel: location-specific impact and an estimated recovery time.
- Post-incident reviews: we write one for every incident with significant customer impact. Ask in your support channel for the summary.
Internally we keep runbooks for each failure mode (region outage, internet outage, server crash, physical disaster). They are not public; summaries are available on request for procurement and security reviews.
Reporting and questions
For a vendor security questionnaire, or copies of our providers' certifications and reports (AWS, Stripe, Twilio), email security@rentaltide.com. RentalTide does not yet have a SOC 2 report of its own.
For technical questions about a specific failure mode, reach out in your support channel and we will walk through it with you.

