kante;11019883 wrote:If a physical webserver with two DNS primary and secondary.
If the secondary is unavailable but the primary is up
is it possible that from some geographical position the server won't be available at all?
Wait a minute; since you seem to be talking about inbound connections to your webserver, I'm guessing you meant name servers here - not "DNS [servers]." The latter would be used by your webserver (or your desktop PC) when it needs to make an outbound connection.
If nothing has changed recently (e.g. the name servers that are authoritative for your domain, your domain's A/CNAME records, etc.), then all of the root NSs should all know about your two NSs, thus they should be handing out both to any client that queries for your domain. Now, whether that client plays by the rules and actually tries both NSs in case one is failed is something you'd have to take up with the client - but for the sake of argument, you can probably assume this will happen.
kante;11019883 wrote:Or always anywhere if primary NS is up server can be considered up?
All of my above reply applies here. Also, I don't think there is such a thing as a "primary" or "secondary" name server. When you execute a query for a domain and the DNS server doesn't have a non-authoritative answer cached and isn't going to do a proxied lookup for you, then it's just going to get the list of NSs from a root server and spit them out at you. There are no priorities assigned to the entries, and I don't think it's in the RFC spec that order matters.
kante;11019883 wrote:I'm asking this coz lately my server provider rebooted the physical machine and the secondary Ns didn't go up
Shouldn't be an issue, since best practice dictates that you have multiple name servers with at least one being on an entirely different subnet than the other(s) (and different physical machines/locations, of course).
kante;11019883 wrote:a collegue of mine had troubles in connection to the webserver via FTP or HTTP...
You'd have to have your colleague define "troubles in connection" before we can even begin to guess whether DNS was at fault here. Otherwise, that description is about as helpful as saying that the problem occurred on a day of the week that ends in the letter "y". Did (s)he try pinging the hostname(s) to see if it/they were being resolved to IP addresses? If so, that pretty much eliminates DNS from being the culprit... unless, of course, the wrong IP was being returned (which might suggest that your name servers aren't replicating zone file changes amongst themselves).
kante;11020157 wrote:What I know and I was kind of sure is:
there are two DNS resolver in case one goes down the other will be available to resolve your names.
Again, I'm getting confused about which end of the spectrum you're referencing here. "DNS resolvers" are things that would run on the client (for example, your desktop PC) that attempt to resolve a hostname into an IP address. A name server is the thing that would get queried in that process and would be responsible for processing the zone file for the domain of the requested hostname and figuring out what response should be given.
kante;11020157 wrote:As it happens under windows. You can specify 2 DNS in network settings. If one of those is down other makes the job.
You can specify anywhere from zero to (more than two - I'm not sure what the actual limit is).
kante;11020157 wrote:But after the situation above happened where this collegue located in a different geographical place (He said he was tryin from different connections even from mobile), but he wasn't not either so far from me (same country),
wasn't able to see websites on the server at the same moment the secondary NS was down
Again, "wasn't able to see websites" is diagnostically useless in troubleshooting the root cause. For example, one possible scenario is that there were routing issues that were preventing his connection from reaching the webserver even though DNS was operating properly. In fact, the same could be true for the connection to the NSs as well (although this is probably unlikely, especially if the "best practice" I mentioned above was followed - that is, after all, the whole reason for it 🙂).