DNS server doesn't deal with subdomain delegations

[Evan Broder]

xvm.mit.edu is configured to delegate dev.xvm.mit.edu to the dev cluster (via use of the override zone files). It serves the NS record correctly, but fails to return that delegation when requesting a subdomain of dev.xvm.mit.edu, instead returning an NXDOMAIN. Consequently a normal DNS query for a dev machine fails with NXDOMAIN.

Demonstration of problem:

white-light$ dig jaunty-test.dev.xvm.mit.edu @18.72.0.3

; <<>> DiG 9.3.6-P1 <<>> jaunty-test.dev.xvm.mit.edu @18.72.0.3
;; global options: printcmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NXDOMAIN, id: 47
;; flags: qr aa rd ra; QUERY: 1, ANSWER: 0, AUTHORITY: 0, ADDITIONAL: 0

;; QUESTION SECTION:
;jaunty-test.dev.xvm.mit.edu. IN A

;; Query time: 14 msec
;; SERVER: 18.72.0.3#53(18.72.0.3)
;; WHEN: Sun Sep 20 18:07:34 2009
;; MSG SIZE rcvd: 45

Details showing what's going wrong:

dhcp-18-111-24-105:~ evan$ dig +nocmd +nostats +norecurs @xvm.mit.edu dev.xvm.mit.edu NS
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 28673
;; flags: qr aa; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 0

;; QUESTION SECTION:
;dev.xvm.mit.edu. IN NS

;; ANSWER SECTION:
dev.xvm.mit.edu. 10800 IN NS xvm-dev.mit.edu.

dhcp-18-111-24-105:~ evan$ dig +nocmd +nostats +norecurs @xvm.mit.edu jaunty-test.dev.xvm.mit.edu
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NXDOMAIN, id: 23668
;; flags: qr; QUERY: 1, ANSWER: 0, AUTHORITY: 0, ADDITIONAL: 0

;; QUESTION SECTION:
;jaunty-test.dev.xvm.mit.edu. IN A

Compare this to what bitsy returns for a request for a name that it's delegating:

dhcp-18-111-24-105:~ evan$ dig +nocmd +nostats +norecurs @bitsy.mit.edu remus.xvm.mit.edu
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 13399
;; flags: qr ra; QUERY: 1, ANSWER: 0, AUTHORITY: 1, ADDITIONAL: 1

;; QUESTION SECTION:
;remus.xvm.mit.edu. IN A

;; AUTHORITY SECTION:
xvm.mit.edu. 21600 IN NS NS1.xvm.mit.edu.

;; ADDITIONAL SECTION:
NS1.xvm.mit.edu. 21600 IN A 18.181.0.62

[Mitchell Berger (mitchb)]

This is now fixed in 854516db26da53a0eda7d28e81084115e1a1a6cc and
running in prod. Finally - a dev cluster that works!