Insights / Platform
Registry Failover Architecture: A Guide for TLD Operators
Originally published on dns.africa
A top-level domain is public infrastructure. When a registry fails, registrars cannot provision or renew names, and if DNS publication is affected, websites and email across the namespace can be at risk. Failover architecture is how a registry operator keeps those consequences contained when hardware, networks, data centres or people let it down.
This guide covers how to design registry failover around recovery objectives, the system of record, registrar access, DNS publication, security and testing, with attention to what operators need to see during an incident.
Start with registry recovery objectives
Every failover design should begin with two numbers for each service: the recovery time objective (how long the service can be unavailable) and the recovery point objective (how much data loss is acceptable). A registry runs several services with different tolerances, so set objectives for each one separately:
- DNS resolution of the TLD zone, which registrants depend on continuously
- EPP provisioning for registrars
- RDAP and WHOIS lookups
- Registrar and operator interfaces, reporting and billing
- Data escrow deposits
gTLD operators must also meet the service levels in their registry agreements, and many ccTLD managers have equivalent commitments to governments or local communities. Those obligations set the floor, and your design should aim above it.
Build around the registry system of record
The registry database is the system of record for every domain, contact, host and DNSSEC record in the namespace. Failover design depends on keeping that data consistent across sites. Replication between primary and secondary environments must be monitored, with alerts when lag grows, and the secondary must be able to take over without corrupting transactions in flight.
The platform matters here. rEngin is the core of the DNS Business registry system, and the failover approach for any registry built on it starts with protecting that core data. Whatever platform you run, the principle is the same: know where the authoritative data lives, how it is replicated, and how you prove both copies match after a failover. Our domain registry architecture guide explains the wider structure.
Design EPP and registrar access for continuity
Registrars connect through EPP, usually with persistent sessions, client certificates and IP allow-lists. During failover, all of that has to keep working. Plan how registrars reach the secondary site: DNS-based switching, shared addresses or published alternative endpoints, and make sure allow-lists and certificates are valid at both locations.
Registrars also need to handle the transition gracefully. Document what happens to open sessions, how in-flight commands are treated and how registrars should reconcile their state afterwards. Clear communication during and after an event protects registrar confidence as much as the technical recovery does. Our guide to EPP vs REST registry APIs (EPP vs REST: Choosing a Registry API Registrars Will Actually Adopt) covers the interface design that makes this easier.
Keep DNS publication independent and verifiable
For registrants, DNS resolution is what matters most. A well-designed registry separates zone publication from provisioning, so that an outage in the registry database does not take the TLD offline. Authoritative name servers should be distributed across networks and locations, often with anycast and more than one provider.
Verification is essential. After any failover, confirm that the zone being published matches the registry database, that serial numbers move forward correctly and that DNSSEC signatures remain valid. A signing failure can be as damaging as an outage, so key management and signing infrastructure need their own redundancy and monitoring.
Choose a failover model that fits the TLD
There is no single correct model. The common options trade cost against recovery speed:
- Active-passive: a warm standby site takes over when the primary fails. Simpler to run, with a short interruption during switchover.
- Active-active: both sites serve traffic, giving faster recovery at the cost of more complex data consistency.
- Cloud or hybrid standby: secondary capacity in a different hosting environment, which can lower cost but needs careful attention to data locality and security.
A small ccTLD may reasonably choose active-passive with strong DNS redundancy, while a high-volume gTLD may justify more. The right choice follows your recovery objectives and your team’s ability to operate the design. For the broader decision about who runs the back-end, see our guide to choosing a shared, managed or self-hosted registry (Shared Registry System, Managed or Self-Hosted? How TLD Operators Should Choose a Registry Back-End).
Plan for the worst case: backups and data escrow
Replication protects against the loss of a site, but it also copies mistakes. If corrupted data, a faulty deployment or a malicious change is replicated to the secondary environment, both copies are damaged. Point-in-time backups, stored separately from the replicated environment and tested by restoring them, are the defence against that scenario.
Data escrow sits one level further out. gTLD registries deposit registry data with an approved escrow agent so the namespace can be recovered even if the operator itself fails, and many ccTLDs follow similar practice. Treat escrow deposits as part of your recovery design: check that they complete on schedule, that they validate, and that the data in them matches the live registry.
Document the order in which you would use each layer. A typical sequence starts with automatic failover to the secondary site, moves to restoring from point-in-time backups if replicated data is unreliable, and keeps escrow as the last resort. Knowing which layer to reach for, and who decides, saves time when it matters most.
Secure the recovery path
Recovery is a moment of elevated risk. Staff work under pressure, normal procedures are bypassed and secondary systems may not receive the same attention as primary ones. Attackers know this, and mistakes made during recovery can outlast the original incident.
Treat the secondary environment as production. Patch it, monitor it and apply the same access controls. Keep privileged credentials for failover in secure, tested storage, and require the same approvals for emergency changes that you would for planned ones, with a faster path that still leaves a record.
Security and compliance are continuous, not tied to a single event. Logging, audit trails and data protection obligations apply during an incident just as they do on an ordinary day, and regulators or contracting parties may ask for evidence afterwards. Our article on what makes registry software compliant covers these requirements, and registry lock adds protection for critical names.
Operator visibility during an incident
During an incident, operators need to understand the state of the registry quickly: which services are affected, which registrars are impacted and whether domain data is consistent. Decisions about switching sites, communicating with registrars or pausing operations depend on that picture.
In rEngin, rEngin Console is where registry operators manage registrars, domains and registry functions, so it is the natural working surface for the operations team before, during and after an event. Make sure your incident runbooks name the tools and views your team will rely on, and that operators can reach them from the secondary environment.
Communication is part of visibility. Prepare templates for registrar notices, decide who approves them, and agree in advance how often updates will be sent during a prolonged incident.
Test failover as an operating discipline
A failover plan that has never been exercised is an assumption. Test regularly, and test realistically: switch real services to the secondary site under controlled conditions, measure recovery against your objectives and record what went wrong.
- Run tabletop exercises for scenarios such as data centre loss, database corruption, DNSSEC signing failure and compromised credentials.
- Schedule live switchovers with advance notice to registrars.
- Verify data consistency and zone correctness after every test.
- Update runbooks and train new staff after each exercise.
Failover testing should also be part of any platform change. If you are planning a move to a new back-end, include failover validation in your registry migration plan before cutover, not after. For the operational differences between country-code and generic namespaces, see our ccTLD vs gTLD operations guide (ccTLD vs gTLD Operations: Running a Registry for Either Model).
Resilience is earned through repetition. The DNS Business team has built and tested failover for registry environments, and the operators who recover calmly are almost always the ones who rehearsed.
Talk through your failover design
If you are reviewing your registry’s resilience, we can walk you through rEngin and how rEngin Console supports your operations team. Book a rEngin demo.