Hello. Welcome to BIND-NANOG, the mailing list where North American BIND
problems (rather than BIND problems in general) are discussed among network
operators. If you would prefer to just discuss general network operations
issues, sorry, there is no such mailing list. If you would prefer to just
discuss BIND problems, see usenet:comp.protocols.dns.bind. Now for our topic:
With the root servers blowing up, someone down here told me that our BIND
8.1 server was looking up less names than other servers. Of course I
denied it, but he coaxed me into running a few tests.
I tested 3 servers, a BIND 8.1 server, a BIND 4.9.5 p1 server, a BIND
4.9.5 server, and a BIND 4.9.3 server.
I tested a total of 100 uncached domains.
16% of the domains couldn't be looked up by the BIND 4.9.3 server
16% of the domains couldn't be looked up by the BIND 4.9.5p1 server
17% of the domains couldn't be looked up by the BIND 4.9.5 server
26% of the domains couldn't be looked up by the BIND 8.1 server
All queries were sent to all the nameservers at the same time. I'm not
sure if this is by chance, but it seems that BIND 8.1 isn't doing as good
a job as the older 4.9.x servers.
Can anyone confirm this?
Without knowing exactly what test you ran, I suspect that noone will be able
to confirm it. However, I can say that BIND-8.1.1 is less promiscuous than
its predecessors, and so it is likely to actually reflect underlying origin
server problems more directly than its predecessors. (This is part of the
cost of preventing AlterNIC, et al, from polluting your cache.) If you would
also run your test (whatever it is) against BIND-4.9.6 I suspect that the
results will be similar to BIND-8.1.1 (which it's not clear that you used;
BIND-8.1 has already been deprecated by its first patch release.)
We now return you to NANOG-FIBRE, wherein the world wide fibre situation and
its possible effects on North American network operations are discussed hourly.