The simplest way to do it is just to create multiple A records. The first A record will be the primary server, the others will be the backups. Most modern browsers should attempt the subsequent IP addresses if the first one fails.
This is a bit like a round-robin DNS but without the IP addresses being rotated for load balancing. Normally for round-robin and similar DDNS setups, you use a short TTL, otherwise you can't change the IP visitors go to when the main server is detected to be down. But if you're not using DDNS, and you just have a fixed list of primary, secondary, tertiary, etc. servers you want clients to try, then, yes, just set a long TTL as you'd normally do for optimal DNS efficiency.
Note: the drawback of this simple method is that, if your primary server is down, the client will still attempt to connect to the primary server first. It will only try to connect to the second server once the first connection fails/times out. And if the second server is also down, then that'll add a delay too. This can make the page load very slow initially (until the host IP of the successful connection is cached). This is why there are more advanced DNS/DDNS failover services out there, where they actively monitor your servers to see if one is down, and modify your DNS records on the fly as appropriate.
For more advanced failover techniques, you should use a specialized DNS hosting service or try cloud hosting or CDNs, the latter of which you don't actually have to do anything yourself, as those infrastructures are designed to be redundant and provide failover protection.
