I set up a VPS (Ubuntu 22.04, at Contabo in Germany) yesterday. Unfortunately, I got the reverse DNS entry incorrect (I entered example.org
instead of example.uk
). I spotted this quickly, and corrected it, and both mxtoolbox.com
and dig
on my local system then showed it correctly resolving to example.uk
.
This new system has to talk to a mail server, which is also Ubuntu 22.04 (at Ionos in the UK). However, postfix
on the mail server rejects incoming mails, because the incoming IP address is still resolving to example.org
. This has been carrying on for about 16 hours. There was a period last night when mails were being delivered, but they're not being delivered now.
On the mail server, dig +noall +answer -x ip-addr
shows a 2-hour countdown in the TTL. However, it then just cycles again, but still shows example.org
. If I flush the local cache with sudo resolvectl flush-caches
it makes no difference, but starts the countdown again. I've tried this on another VPS in the same datacentre (Ionos, UK) and it shows the same results.
So, I'm guessing that Ionos has an rDNS cache that ignores the upstream TTL (since mxtoolbox
and my own local dig
have been showing the correct example.uk
result for 16+ hours). Does that make sense? Can I set up the local resolver on the mail server to (temporarily) ignore any upstream caches and get an authoritative answer? How do you even get an authoritative answer on reverse DNS?