[Unbound-users] Strange SERVFAIL from unbound

[Florian Weimer]

* W. C. A. Wijngaards:

> Looked at zen.spamhaus.org; but I could not see how harden-glue would
> make an impact. Or how A records could time out while the AAAA do not.
> They have the same timeout value (4 hours).

Are you sure they are cached at the same time?

We had lots of fun with additional section processing for MX records
and A and AAAA records being cached at different times.  It meant that
some of the time, only AAAA records ended up in the additional
section.  Exim assumed that additional section processing has been
performed properly (because the AAAA RRset was present), and if the
host had no working IPv6 connection, it could not deliver the message.

In response to that, Exim was changed not to look at the additional
section.  Considering the cache manipulation risk, I think additional
section processing is a failed approach anyway.