timeout of glue A record

[Gunnar Lindberg <lindberg@cdg.chalmers.se>]

Is this a bug or a flaw in the DNS design? Some days ago we saw:

    % dig @our-ns1 foo.bar ns
    ;; ANSWER SECTION:
    foo.bar.            23h19m44s IN NS  b.ns.foo.bar.
    foo.bar.            23h19m44s IN NS  a.ns.foo.bar.
    ;; ADDITIONAL SECTION:
    ;; None - NB  <-----------------

    % dig @our-ns2 foo.bar ns
    ;; ANSWER SECTION:
    foo.bar.            2m51s IN NS     a.ns.foo.bar.
    foo.bar.            2m51s IN NS     b.ns.foo.bar.
    ;; ADDITIONAL SECTION:
    a.ns.foo.bar.       2m51s IN A      192.n.m.1
    b.ns.foo.bar.       2m51s IN A      192.n.m.2

So, our-ns1 had a reference to their NSs (the names to be specific)
but the corresponding glue A records were missing. What happened
before to get us into this state is beyond my knowledge, but since
our-ns2 was counting down the glue A's ttl I take it they reached 0
and were simply discarded.

After 2m51s our-ns2 was able to refresh both set of records, while
our-ns1 stayed out of sync for another 23h19m44s.

It would seem reasonable to me that the DNS should not keep NS
records unless it has at least some A records to support them.

Is this a bug or design flaw?

	Gunnar Lindberg
#----------------------------------------------------------------------
# To unsubscribe, send a message to <dnsop-request@cafax.se>.