[bind10-dev] using our resolver for real work (and NSAS failures)

[Jeremy C. Reed]

The old git server used our resolver for over a year. I have been using 
it as a forwarder for over a year.  Over the past couple months, I 
configured most of our build systems, the BIND 10 web server, and main 
BIND 10 zone DNS server to only use our resolver. (The open resolvers 
running on both n10 and git boxes.)

A few times I have noticed problems.

I noticed yesterday that sendmail had problems sending mail from the 
Trac server to one of the bug submitters (not part of the team) due to 
DNS lookup failures.

I opened a few tickets about some problems:

NSAS permanently cache unreachable namservers?
http://bind10.isc.org/ticket/2031

NSAS wrongly recording nameservers are unreachable
http://bind10.isc.org/ticket/2030

NSAS repeating same tasks near simultaneously
http://bind10.isc.org/ticket/2029

Before I opened the tickets, one of our developers via jabber said: 
``I'm afraid we won't have time to fix the current b10-resolver ... 
except for very critical bug fixes like reproducible crash or security 
bugs.'' He suggested we ``keep [a] record [of] these weird things so we 
can do it right in our "real resolver" work later this year ... 
[because] trying to fix these for the current implementation will be a 
duplicate effort.''

I think my problems are related to my the old, now-closed (but not fixed 
as far as I know), ticket 833:

[b10-resolver] Nameservers unreachable but really are
http://bind10.isc.org/ticket/833

The resolver has been in our tree for a couple years. Does anyone else 
use it?


The BIND 10 systems are still using our resolver. I worked around one 
problem by restarting b10-resolver.