I have setup DNS entries so that a certain FQDN has four A records, with four IPv4 addresses of four networks boxes converging thru a router to a single https server. That does load balancing across the boxes.
I'm worried about what happens if a network box goes down. The DNS resolution mechanism will return IPv4 addresses even if the link is down. I read here that (trimmed)
pretty much every browser receive the full list of A records, and check others if the one it is using fails.
which is fine. But my server provides a web service: it's clients are typically not web browsers. I know one uses libcurl under Linux Centos 7, but there could be others.
How would various https stacks (including but not limited to the above) handle a failed link (as in network box powered down or its line severed)?
Any application querying DNS records receives all the results. How they are handled depends on the application, it can try the first, a fixed number, or all of them.
Most applications should try all records, but if it is important to you, you should verify yourself. Just setup a DNS name that points to some addresses that don't respond, then try to access that name and see how many of the addresses are tried, andhow long the timeout is.