Configuring DNS failover

What DNS failover is#

DNS failover means automatically switching users to a backup server when the primary goes down. In an ideal setup, a health check detects failure, DNS stops returning the failed server’s IP, and clients resolve to a healthy backup. Users get continuity without manual intervention.

Edge DNS geo-routing#

Edge DNS supports geo-routing: routing different users to different servers based on their location. This works today. For example, you can return IP A for users in Europe and IP B for users in the Americas. This is useful for multi-region deployments where you want users to hit the nearest origin.

Configure geo-routing in the console or via the API. Add A (or AAAA) records with geographic filters and specify which regions resolve to which IPs.

Multiple A records for round-robin#

You can add multiple A records for the same hostname. Resolvers typically round-robin between them. This distributes load across several servers but does not provide automatic failover: if one IP is down, clients may still be directed to it. DNS does not remove unhealthy IPs automatically.

Monitoring DNS with Edge metrics#

Edge provides DNS metrics: query volume, response times, and resolution data. Use these to monitor DNS health and spot anomalies. They complement application-level monitoring but do not replace health checks for failover logic.

Health-check-based failover: not yet implemented#

Workarounds#

• External monitoring and API updates: use an external monitoring service (e.g. UptimeRobot, Pingdom) to check your endpoints. When a failure is detected, call the Edge API to update DNS records, for example to remove the failed IP or add a CNAME to a backup. Requires automation (a script or Lambda) and API access.
• CDN as a resilient layer: put Edge CDN in front of your origin. The CDN caches static content and can fail over to a backup origin if configured. For dynamic content, origin health still matters, but the CDN adds redundancy for cached responses.
• Manual failover: the console and CLI always allow manual record updates. For planned maintenance or known outages, switch records manually. It isn’t automated, but it’s reliable.