bind 9.3.1 still causes delays in multiples of 2 seconds

[Mark Andrews]

> Hi everyone,
> 
> When asking a bind-9.3.1 to resolve something that's not in its cache, 
> it will sit idle for 2 seconds (or multiples thereof) before asking 
> asking some of the nameservers. But from prior discussions in this 
> newsgroup, I thought this particular bug was fixed in the 9.3.1-release?
> 
> Packet trace: (timestamps are relative)
>   client -> ns1 DNS C www.sunray.nl. Internet Addr ?
>   0.00350 ns1 -> ns.nic.nl DNS C www.sunray.nl. Internet Addr ?
>   0.00086 ns.nic.nl -> ns1 DNS R
> (time passes.. nothing happens)
>   2.01631 ns1 -> 213.136.12.59 DNS C www.sunray.nl. Internet Addr ?
>   0.00035 213.136.12.59 -> ns1 DNS R www.sunray.nl. Internet Addr 
> 213.136.0.188
>   0.00088 ns1 -> client DNS R www.sunray.nl. Internet Addr 213.136.0.188
> 
> Or, as dig puts it:  'Total query time: 2018 msec'
> 
> Can anyone confirm these findings?
> 
> I'm running Bind on Solaris 8/9. Compiling bind with '--disable-ipv6' is 
> not really an option for us as some of our nameservers do run IPv6.
> For now I guess I'll have to disable IPv6 support by starting bind with 
> the '-4' options on all machines that don't have IPv6 connectivity.
> 
> Regards, Paul Boven.
> .

	Look at lib/dns/resolver.c:resquery_senddone().  There may
	be another error code that needs handling.  Currently we
	handle the following with a fast retry.

                case ISC_R_HOSTUNREACH:
                case ISC_R_NETUNREACH:
                case ISC_R_NOPERM:
                case ISC_R_ADDRNOTAVAIL:
                case ISC_R_CONNREFUSED:

	Mark
--
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark_Andrews at isc.org