Dns sometimes fails using Google DNS / automatic dnssec


We've been seeing automatic RRSIG records on Google DNS lately, the en They are not always provided. They cause problems for some of our customers in a weird way I cannot explain. For them these records do not resolve but I cannot reproduce it.

So when I run dig command

dig @ m1.mailplus.nl

it often provides the RRSIG record (but e.g. the TXT record will not be signed). I've heard that DNS may fall back to TCP and/or may be filtered by firewalls if UDP is over 512 bytes. However, the request is not that long, about 200 bytes if I interpret the answer correctly.

Can someone come up with a good explanation why a tiny percentage of our customers cannot resolve (some of) our domains?

Btw, our nameservers (transip.nl) only provide DNSSEC records if explicitly asked. What is standard here?


David Hofstee

Hi, David

I work at Google Public DNS and will take a look at this issue. No
RRSIG should be returned unless the client set the DO bit to ask for


root@e3:/home/services# dig @ m1.mailplus.nl

; <<>> DiG 9.7.3 <<>> @ m1.mailplus.nl
; (1 server found)
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 38880
;; flags: qr rd ra; QUERY: 1, ANSWER: 2, AUTHORITY: 0, ADDITIONAL: 0

;m1.mailplus.nl. IN A

m1.mailplus.nl. 1867 IN A
m1.mailplus.nl. 1867 IN RRSIG A 7 3 3600 20130517082302 20121115082302 3767 mailplus.nl. WzKY2FnTbF8MOhAuDvnrPkpgskeH4aI1YByh6zBX1z1pQRo8YIcxzlSN tHv2LnKUk+0n6iIXqV77sHynHHP/Y/a0bMKYKIDuK8Gtz47AVDJaU0eX 0FR8F5qqw897ClGf5ISa0njPLFVyF/NJ6hNViDYzOhhHGi58dhZmhKWF ujs=

;; Query time: 5 msec
;; WHEN: Thu Nov 15 16:05:26 2012
;; MSG SIZE rcvd: 219

It looks like if the server has the RRSIG RR, it returns it. For example, a
query with +dnssec will cause it to cache the RRSIG, after which it returns
it even if +dnssec not specified.

Hi, we have found the bug that caused this problem. It was introduced
in a very recent release. The fix is on its way.

Thanks very much for the report,

It's weird. If you repeatedly query without the DO bit, you get a
mixture of responses with and without an RRSIG and with varying TTLs. With
DO it appears to consistently return an RRSIG in the answer and the TTL
drops monotonically. is similar except DO=0 replies don't include
RRSIGs. (Querying from JANET UK and hitting some servers a lethargic 12ms

while sleep 1; do dig +dnssec @ m1.mailplus.nl; done