DNS failover can reduce the time users spend facing a website outage, but it is not a complete disaster-recovery plan by itself. This checklist explains how to choose a standby environment, configure health checks and DNS behavior, keep content and data aligned, protect email and SSL, test the process, and return traffic safely when the primary service is restored.
Overview
DNS failover for websites means directing visitors to an alternate destination when the primary host, server, region, or application becomes unavailable. A DNS provider, traffic manager, or CDN typically evaluates the primary endpoint and changes the response when a defined condition is met.
The practical goal is not to make an outage invisible. DNS records are cached by recursive resolvers and browsers, and some users may continue reaching the previous destination until cached answers expire. Failover is therefore a continuity control that can reduce disruption, not a guarantee of instant switching.
Before designing a setup, define the service you are protecting. A brochure site may need a static standby page, while an online store, customer portal, or booking system may require synchronized application files, a consistent database, payment integrations, authentication, and a recovery process for transactions made during the incident. Also decide whether email is part of the recovery scope. Website failover usually does not automatically protect mail delivery, calendars, file storage, or third-party services.
DNS load balancing and DNS failover are related but different. Load balancing distributes normal traffic across multiple healthy endpoints. Failover keeps a primary endpoint in service and uses a standby when the primary fails. Some implementations combine both approaches, but the operating procedures and testing requirements should be documented separately.
For broader continuity planning, pair this work with a documented website backup and recovery plan, a hosting review, and an ownership record for domains and DNS credentials. A redundant host cannot help if the organization cannot access the DNS account during an incident.
Checklist by scenario
Compare the main approaches
| Approach | How it works | Best fit | Important limitation |
|---|---|---|---|
| Active-passive | A primary origin serves traffic while a standby remains ready but receives little or no production traffic. | Business sites that need a simpler recovery path and can tolerate a controlled switch. | The standby can drift unless content, configuration, and dependencies are regularly synchronized. |
| Active-active | Two or more origins serve production traffic at the same time. | Applications designed for multi-region or multi-server operation. | Requires careful session, database, deployment, and monitoring design. |
| DNS health-check failover | A DNS or traffic service checks an endpoint and returns the standby destination when the primary fails. | Sites where the DNS provider supports suitable checks and routing rules. | DNS caching, check intervals, provider limits, and false positives affect behavior. |
| CDN-based failover | A CDN or edge platform routes requests between origins or serves cached or static content when an origin is unavailable. | Sites already using an edge platform, especially those with cacheable content. | Dynamic requests, uncached pages, purges, and origin-specific features still need a recovery plan. |
| Manual failover | An authorized operator changes DNS, CDN, or routing settings after confirming an incident. | Low-traffic or lower-criticality sites where automated routing is not justified. | Recovery depends on people, access, documented steps, and timely detection. |
Step 1: Classify the service and set recovery targets
Write down the acceptable recovery time and the amount of data the business can afford to lose. These targets should be practical rather than aspirational. A marketing site may recover from a recent backup, while a transactional application may need near-current data and a more sophisticated architecture.
- List the domains, subdomains, URL paths, and public IP or hostname destinations in scope.
- Identify which functions are essential: pages, forms, logins, checkout, APIs, uploads, and administrative access.
- Record dependencies such as databases, object storage, payment services, identity providers, webhooks, and external APIs.
- Assign an incident owner and a backup operator with access to the DNS, hosting, CDN, and monitoring accounts.
Step 2: Select and prepare the standby origin
Choose a standby host or region that can serve the required version of the application. It might be a second hosting environment, a separate cloud region, a static site location, or a CDN-supported origin. The standby should have the same domain configuration, runtime assumptions, environment variables, security controls, and deployment process as the primary wherever possible.
Do not treat a server that has never been restored as a standby. Deploy the application, install the required certificates, configure redirects, verify file permissions, and confirm that the standby can answer requests using the intended hostname. Review the options in business web hosting plans compared if the recovery design requires a change in hosting capacity.
Step 3: Configure health checks and DNS behavior
Create a health-check endpoint that represents real service health. A response showing only that the web server is running may miss a failed database, broken application dependency, or unusable deployment. At the same time, a check that depends on too many external systems can trigger unnecessary failovers. Use a lightweight endpoint with clear success criteria, and test it from more than one monitoring location if the provider supports that.
Document the check path, expected status, timeout, retry behavior, failure threshold, recovery threshold, and notification destination. Understand whether the provider checks the hostname, a specific IP, an HTTP response, HTTPS certificate behavior, or a TCP port. Provider limits and available routing features vary, so confirm the current configuration in the relevant product documentation before relying on it.
Set the planned DNS TTL according to the provider's capabilities and the business tolerance for cached answers. A shorter TTL may help planned changes take effect sooner, but it cannot force every resolver to discard an existing answer immediately. Avoid changing TTL for the first time during an outage.
Step 4: Synchronize content, databases, and configuration
Define how each type of data reaches the standby. Code can be deployed through the same release pipeline; uploaded files may require replication or a shared storage service; databases may need replication, scheduled exports, or a deliberate read-only recovery mode. Decide which system is authoritative during a failover and how writes are handled to avoid conflicting data.
- Test that the standby has the current application version and configuration.
- Confirm that media, downloads, and user uploads are available.
- Document database replication status, backup points, credentials, and promotion steps.
- Decide whether forms, orders, comments, and account changes are disabled, queued, or accepted during recovery.
- Check third-party allowlists, webhook destinations, API keys, and payment settings.
Step 5: Protect HTTPS and email
Every hostname used after failover must have a valid SSL certificate and a compatible TLS configuration. Check the certificate before the incident, including alternate names for subdomains. The related guide on setting up SSL and HTTPS without breaking redirects can help structure this review.
Keep website DNS records separate from mail records in the runbook. MX, SPF, DKIM, and DMARC records should not be changed casually as part of a web failover. Confirm that the standby application sends mail through an approved provider and that reply addresses, tracking links, and notification workflows remain valid.
Step 6: Test, declare, and roll back
Run a controlled failover in a maintenance window. Record the time the primary was disabled, the time monitoring detected it, the time routing changed, and the time users could complete an essential transaction. Test from multiple networks, including mobile access, and verify both cached and uncached pages.
When the primary is stable, do not immediately switch everyone back. Confirm application health, data consistency, deployment version, and monitoring first. Restore traffic in a controlled manner if the platform allows it, then watch logs, error rates, transactions, and support reports. Record any data created during failover and reconcile it before accepting normal writes on the primary.
What to double-check
- DNS scope: Confirm whether the setup covers the apex domain, www host, application subdomains, API endpoints, and any regional names.
- Cache behavior: Review CDN caching, browser caching, stale content rules, purge controls, and whether errors can be cached.
- Health-check realism: Verify that a passing check means users can perform the required action, not merely that a port is open.
- Access and security: Store DNS and hosting recovery credentials securely, use separate operator accounts where available, and require an approval path for emergency changes.
- Backups: Confirm that backups can be restored into the standby environment and that restoration steps are written down.
- Observability: Ensure alerts cover DNS changes, origin health, certificate expiry, database replication, and important user journeys.
- Provider behavior: Recheck health-check limits, failover routing, propagation guidance, CDN origin behavior, support procedures, and pricing before a major continuity exercise.
A DNS record change is also an account and domain-management event. Keep ownership details, renewal dates, and recovery contacts current by following a documented domain portfolio management process.
Common mistakes
Assuming DNS failover is instant. Cached answers can continue directing some visitors to the primary. Plan for a transition period and communicate what users should try if they receive an old destination.
Using a standby that has never been exercised. An untested server may lack certificates, recent uploads, environment variables, firewall rules, or a working database connection. A scheduled restore and failover drill is more useful than an unverified architecture diagram.
Failing over the website but not the data. Sending requests to an alternate application does not solve database loss, conflicting writes, or missing files. Define data behavior before choosing routing technology.
Making the health check too simple or too sensitive. A check that only tests a web server can miss application failure; one that fails for a minor dependency can cause an unnecessary switch. Tune thresholds using observed behavior.
Forgetting rollback. Recovery is incomplete if the team cannot safely return to the primary. Document who authorizes rollback, how data is reconciled, how traffic is verified, and when the incident is closed.
Changing DNS during a domain or hosting transfer without records. Export or document the full zone, including verification records and mail settings, before making major account or registrar changes. The business domain transfer checklist provides a useful companion process.
When to revisit
Review this failover plan before seasonal planning cycles, major campaigns, product launches, hosting changes, domain transfers, application migrations, and changes to payment or identity providers. Also revisit it whenever the DNS provider, CDN, hosting platform, health-check features, propagation guidance, or pricing changes. These details can change independently of the website itself.
At least twice a year, run a documented continuity exercise. Start with a review of contacts and credentials, then restore a current backup, test the standby, simulate a primary outage, validate DNS and HTTPS, check essential user journeys, and practice rollback. Afterward, record the actual detection and recovery times, failed steps, missing access, and any data reconciliation work.
Use this short runbook for the next exercise:
- Confirm the incident or test window and assign an incident lead.
- Capture the current DNS, CDN, application, database, and backup state.
- Place the standby in service according to the documented routing procedure.
- Verify DNS responses, HTTPS, pages, forms, logins, transactions, email notifications, and monitoring.
- Communicate the status and record timestamps and observed user impact.
- Restore the primary only after its data and application state are confirmed.
- Roll traffic back gradually, reconcile any writes, and monitor for regressions.
- Update the checklist, architecture notes, and ownership records based on what the drill revealed.
A failover design earns its place in a business continuity plan when the team can operate it under pressure, not merely when the DNS rules appear correct. Keep the procedure accessible, test it after infrastructure changes, and treat every exercise as an opportunity to simplify recovery.