[Pdns-users] Recurring SERVFAIL problem for api.met.no

[sthaug at nethelp.no]

Environment: FreeBSD 11.2-STABLE, PowerDNS Recursor 4.1.12, dnssec=log-fail

I have a recurring problem with the domain name "api.met.no". Normally
it resolves like this:

api.met.no.             3600    IN      CNAME   external.api.met.no.
external.api.met.no.    3600    IN      CNAME   os-157-249-177-128.ares.met.no.
os-157-249-177-128.ares.met.no. 3600 IN A       157.249.177.128

;; AUTHORITY SECTION:
ares.met.no.            3600    IN      NS      dns10.met.no.
ares.met.no.            3600    IN      NS      dns11.met.no.

;; ADDITIONAL SECTION:
dns10.met.no.           3600    IN      A       157.249.32.14
dns11.met.no.           3600    IN      A       157.249.32.15

and PowerDNS Recursor has no problems resolving the name.

Sometimes PowerDNS Recursor gets into a state where this name results
in SERVFAIL - and in this case a "rec_control wipe-cache met.no$" will
*not* fix the problem. A restart of the PowerDNS Recursor fixes the
problem.

I have copious output from "rec_control trace met.no" but I'm unable
to determine directly from this output what the problem is. I'm also
somewhat hesitant to post the whole trace output here, since it is
rather voluminous.

Any suggestions on how to process with troubleshooting here?

Steinar Haug, AS2116