We have a three nameservers ns1.ukit.com ns2.ukit.com and ns3.ukit.com.
The software is PowerDNS 4.1.5 packed into docker containers, the network ports are exposed via docker-proxy (not "--net=host").
The backend for the PowerDNS is "remote", and it interacts with our self-made restful service which serves records from MongoDB.
The problem is that Google's public DNS sometimes returns our SOA record for the queries of A-records, like there is no A-record at all. In result our websites are becoming inaccessible at random time from random locations across the world, according to reports from host-tracker.com (a geo-spread monitoring service) our users also have reported us about theese problems with availability of their sites. We also could get these empty anwers ourselves quering 22.214.171.124 and 126.96.36.199 (approximately 1 from 10 attempts), while other public DNS services seemed not to encountered the same, we tried 188.8.131.52 and 184.108.40.206.
The questions are:
- Why Google's public DNS does so?
- Are there any special requirements for authoritative DNS-servers to be operable normally via Google's public DNS? E.g. response time, answers format etc.
- Is there a way to check compatibility with Google's public DNS ourselves, apart from listed here https://developers.google.com/speed/public-dns/docs/troubleshooting?
The date and time you encountered the problem
Started from 01.30.2019 15-00 UTC.
ns1 - Moscow, Russia
ns2 - Ashburn, USA
ns3 - Amsterdam, Netherlands
The platform on which you are noticing the problem (e.g. Mac, Windows, router, etc.)
I think it doesn't depend on platform
The hostname(s) for which you are having a problem
ukit.com, ulanding.io, divly.ru (actually there are some more).
Whether the problem is continuous or intermittent
I'd say it's continiously intermittent.
The links to the tools' name server diagnosis report page
The output of the commands you ran in the diagnostic tests