Floating IPs are reassignable; point the IP to a new server when failing over or migrating. Reduces DNS TTL dependency. Not all providers offer them; check support and any ARP/announce delay.
What floating IPs are
- Reassignable: You own (or lease) an IP that can be moved from one server (or NIC) to another without changing DNS. The same IP is now served by a different machine.
- Use cases: Failover (primary dies, point the IP to the standby); migration (new server is ready, move the IP from old to new); maintenance (drain one node, move IP to another).
- vs DNS: DNS failover requires TTL to expire and clients to re-resolve. Floating IP change is at the network layer; existing connections may still break but new connections go to the new server immediately (after ARP/announce).
How it works
- Provider control: You call the provider API or use their UI to reassign the floating IP from server A to server B. The provider updates routing/ARP so the IP is reachable on the new host.
- Delay: There can be a short period (seconds) where the IP is not reachable or traffic is still going to the old host. Ask the provider about typical delay and any best practices (e.g. gratuitous ARP, BGP if applicable).
- Health checks: Automate failover by having a script or orchestrator that detects failure and reassigns the floating IP to the standby. Test the procedure.
Limitations
- Not universal: Not every host offers floating IPs. Some offer them only in certain regions or products (e.g. dedicated, not VPS). Check before you design for it.
- Single point: The floating IP itself is a single logical endpoint; if the provider has an issue with that IP or its routing, you are affected. Combine with DNS for redundancy where needed.
Summary
Floating IPs let you move an IP between servers for failover or migration without waiting for DNS. Check provider support and ARP/announce delay; automate failover where possible.
