"DNS propagation" n'est pas a real phenomenon, per se. Rather, it is le manifest effect of le caching functionality specified in le DNS protocol. Saying that changes "propagate" entre DNS servers is a convenient falsehood that's, arguably, easier to explain to non-technical users than describing tous of le details of le DNS protocol. C'est pas really how le protocol works, though.
Recursive DNS servers make queries on behalf of clients. Recursive DNS servers, typically run by ISPs ou IT departments, are used by client computers to resolve names of Internet resources. Recursive DNS servers cache le results of queries they make to improve efficiency. Queries for already-cached information can be answered sans making tout additional queries. The duration, in seconds, that a result is cached is supposed to be based on a configurable value called le Time To Live (TTL). This value is specified by le authoritative DNS server for le record queried.
Il y a no one answer to tous le questions being asked parce que DNS is a distributed protocol. The behavior of DNS depends on le configuration of le authoritative DNS server for a given record, le configuration of recursive DNS servers making queries on behalf of client computers, et DNS caching functionality built-in to le client computers' operating systems.
C'est good practice to specify a TTL value short enough to accommodate neecssary day-to-day changes to DNS records, mais long enough so to créez un "win" in caching (i.e. pas so short as to age-out of cache too quickly to provide tout efficiency improvement). Employing a balanced strategy avec TTL values results in a "win" for everyone. It reduces les deux le load et bandwidth utilization for le authoritative DNS servers for a given domain, le root servers, et le TLD servers. It reduces le upstream bandwidth utilization for le operator of le recursive DNS server. It results in quicker query responses for client computers.
As a DNS record's TTL is set lower load et bandwidth utilization on le authoritative DNS servers will increase parce que recursive DNS servers will pas be able to cache le result for a long duration. As a record's TTL is higher changes to records will pas appear to "take effect" quickly parce que client computers will continue to receive cached results stored on leur recursive DNS servers. Setting le optimal TTL comes down to a balancing act entre utilization et ability to change records quickly et see those changes reflected on clients.
C'est worth noting that certains ISPs are abusive et ignore le TTL values specified by le authoritative DNS servers (substituting leur own administrative override, qui is a violation of RFC). Il y a nothing to be done about this, depuis a technical perspective. If le operators of abusive DNS servers can be located complaints to leur systems administrators might result in leur implementing best practices (arguably what amounts to common sense for tout network engineer familiar avec DNS). This particular type of abuse n'est pas a technical problem.
If everybody "plays by le rules" changes to DNS records can "take effect" très quickly. In le case of changing le IP address assigned to an "A" record, par exemple, an exponential backoff of le TTL value would be performed, leading up to le time le change will be made. The TTL might start at 1 day, par exemple, et be decreased to 12 hours for a 24 hour period, alors 6 hours for a 12 hour period, 3 hours for a 6 hour period, etc, down to certains suitably small interval. Once le TTL has been backed-off le record can be changed et le TTL brought back up to le desired value for day-to-day operations. (C'est pas nécessaire to use an exponential backoff, however this strategy minimizes le time le record will have a low TTL et decreases load on le authoritative DNS server.)
After making a DNS record change logs should be monitored for access attempts being made en conséquence of le old DNS record. In le example of changing an "A" record to refer to a nouveau IP address a server should remain present at le old IP address to handle access attempts resulting depuis client computers encore using le old "A" record. Once access attempts based on le old record have reached an acceptably low level le old IP address can be disused. If le requests related to an old record are pas abating quickly it is possible that (as described above) a recursive DNS server is ignoring le authoritative TTL. Knowing le source IP address of an access attempt, cependant, does pas provide direct information as to le recursive DNS server responsible for supplying an old record. If le IP addresses of errant access attempts are tous related to a single ISP it may be possible to locate le offending DNS server et contact its operator.
Personally, J'ai seen changes "take effect" immediately, in a peu de hours, et dans certains cas avec a particular brain-damaged ISP, après plusieurs days. Doing a backoff of votre TTL et being mindful of how le process works will increase votre changes for success, mais you ne peut pas ever be sure what certains well-meaning idiot might be doing avec leur recursive DNS servers.
