On Thu, 26 Oct 2017 22:42:12 -0400, Ken Hornstein said: > >Thus said Ken Hornstein on Thu, 26 Oct 2017 14:22:37 -0400: > >But the A record for that name is only 1 second... strange: > > > >$ dnsqr a cyrus.andrew.cmu.edu > >1 cyrus.andrew.cmu.edu: > >86 bytes, 1+3+0+0 records, response, noerror > >query: 1 cyrus.andrew.cmu.edu > >answer: cyrus.andrew.cmu.edu 1 A 128.2.158.26 > >answer: cyrus.andrew.cmu.edu 1 A 128.2.105.45 > >answer: cyrus.andrew.cmu.edu 1 A 128.2.158.29 > > > >I wonder what the purpose of such a configuration might be. > > I suspect they wanted to provide high reliability and have clients fall > over to one of the other servers quickly. But even Google has a TTL > of 300 seconds for imap.gmail.com.
This can also happen if you've been slowly ramping down the TTL before an address change - you may have 3 days as the TTL. A week before, set TTL to one day, 2 days before, set to an hour, etc and then for the last hour finally down to some tiny like a few seconds. After the new DNS entry is out, you push out a new one with a more sane TTL. And if you forget that last step, this is what you get.
pgpeCW5TmYClF.pgp
Description: PGP signature
_______________________________________________ Nmh-workers mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/nmh-workers
