DNS failover and health checks
If the server a name points to goes down, that name keeps pointing at it — and your visitors keep hitting a dead address — until something changes the record. DNS failover is that something: it notices the failure and moves the name to a server that is actually up.
A DNS record is just a label pointing at an address. If that address stops answering, the label does not care — it keeps sending everyone there. DNS failover adds a watcher: it health-checks the target, and when the target fails, it rewrites the record to point at a healthy one instead.
Health checks drive it
The failover system probes your servers — a TCP connection, an HTTP request, something that proves the service is really working, not just that the machine pings. When enough checks fail, it swaps the DNS record to a standby. The quality of the check matters: pinging the host tells you the box is on, not that your application is serving. See monitoring your own server.
The TTL is the catch
A DNS answer is cached by resolvers for its TTL, so a failover only takes effect as fast as that TTL allows — and not even then for clients holding an old answer. Keep the TTL low on failover-critical records (say 30–60 seconds) so recovery is quick, accepting slightly more DNS traffic as the price. See choosing DNS TTLs.
Its honest weakness
DNS failover is never instant. Between the failure, the health check confirming it, the record changing, and caches expiring, some users hit the dead server for a while. It is a good, cheap safety net — it is not seamless. Do not sell it as zero-downtime, because the first minutes of an outage will still bite. See what 99.9% actually means.
Where it fits
Failover at the DNS layer is simplest and works across locations. For faster, tighter failover within a location, a floating IP or a load balancer in front of two servers reacts in seconds, not minutes — see high availability, honestly and spreading traffic across two servers. Anycast is the heavy-duty option. Many setups layer them: DNS failover between sites, faster failover within each.
Set it up with us
If you want a standby server and DNS failover between them, tell us the service and what a real health check should test — we would rather it checks the thing users actually use than just whether the server is powered on.
Still stuck? Mail support@novogara.com — an engineer answers, at any hour. Back to the knowledge base