On 15. 10. 20 11:51, Balakrishnan Balasubramanian wrote:
  Thanks for checking! For some domains, 'A'
record works fine but AAAA record
 crashes. May be it has do with some ipv6 issue in my router. 
As far as I can tell from verbose log here
https://debug.knot-resolver.cz/query.py?qname=mail.smtp2go.com&qtype=AA…
this domain does not have AAAA records so the answer seems to be correct.
  Is there a way to enable a backup resolver only for
failed queries? I can see
 policy.FORWARD and policy.TLS_FORWARD functions. But I think they forward all
 queries, not just the failed ones.  
You are right, policy.FORWARD* policies are unconditional. Currently Knot Resolver does
not have this Frankenstein-style feature to combine direct queries with forwarding for a
single request.
  Also is there a way to get all the direct dns queries
(excluding name server
 ones) without turning on full verbose logging? 
I'm not sure what you mean. Maybe
https://knot-resolver.readthedocs.io/en/v5.1.3/modules-policy.html#policy.D…
could help?
Petr Špaček  @  CZ.NIC
 Thanks,
 Bala
 On Thursday, October 15, 2020 3:06:02 AM EDT Vladimír Čunát wrote:
> On 10/14/20 8:48 PM, Balakrishnan Balasubramanian wrote:
>> Thanks! Got verbose logging. But not sure what is the issue. Attaching
>> logs.
> I'm not sure either.  It looks like kresd is doing nothing wrong.  The
> two servers in **.ns.els-gms.att.net. don't appear to reply over UDP at
> all and send garbage over TCP.  When I query the same IP addresses from
> here they reply OK(-ish) and apparently also from other places (like
> from other public resolvers or 
dnsviz.net).
>
> It's possible that something in your network is interfering with the
> queries.  Overall I suspect the cause will be hard to track down.
>
> --Vladimir