How to organize a good resource fault tolerance?

Question

High availability

I asked myself the question of creating the most fault-tolerant system for my web resource. As I understand it, mirroring is necessary in everything to achieve, therefore, I don’t understand how to mirror some things. let's say there is a site, rogaikopyta.rf, how the user accesses the site: request to dns, get ip, request to ip, get content
What exactly I don’t understand:
Request to dns. Obviously, in order to implement mirroring, you need your own NS servers. Obviously, you need two of them, on different providers and, preferably, different physical locations. I understand it. Further, how to make dns check which ip address with the resource is alive and give the right ip address? provided that I have 2 or more copies of the site? There is another problem here, because dns is cached, and if he gives up the IP, and then it falls, will he go looking for another address or not?
Well, in general, if you tell how "adult" sites make fault tolerance, it will be cool.

Answer 1

1. Too broad a topic to discuss within a single issue.
2. It is impossible to give an unambiguous answer to this question.

Answer 2

load balancer
further in the search for Habr and in general
oh, and
even master slave replication

Answer 3

Doing it "in an adult way", they usually announce the same addresses (their own) from several points for the resource. This removes the problem of "the user's DNS has cached an invalid IP address". How it will then be processed inside is another question.