ccu.cc 600 CNAME castor.cc.umanitoba.ca.
600 CNAME pollux.cc.umanitoba.ca.
600 CNAME antares.cc.umanitoba.ca.
600 CNAME toliman.cc.umanitoba.ca.
600 CNAME merak.cc.umanitoba.ca.
When I query a primary or secondary, with the `host' command, I see a
nice rotation with each query. When I do it on a cache-only nameserver,
I get the same result each time. How come?
--
-Gary Mills- -Unix Support- -U of M Computing and Network Systems-
When you query the primary or secondary, I'll bet you only get one CNAME
each time (a different one, due to the rotation).
Well, the same thing happens when the caching server makes its recursive
query. It gets one answer, and puts it in its cache. Then it returns that
answer every time it's queried, until the TTL expires.
--
Barry Margolin
BBN PlaNET Corporation, Cambridge, MA
bar...@bbnplanet.com
Phone (617) 873-3126 - Fax (617) 873-6351
Multiple CNAMEs are a bug. BIND should not be allowing you to enter them
into your zone file. "ccu.cc" is either an alias in which case it has a
single canonical name, or it is not an alias and it is a canonical name.
Though I have used and recommended multiple CNAME RR's in an "RRset" (which
it really isn't) in the past, I hereby apologize and recommend that anyone
now doing it, please stop.
In the above example you should use multiple A RRs.
--
Paul Vixie
La Honda, CA "Illegitimibus non carborundum."
<pa...@vix.com>
pacbell!vixie!paul
I was recently reading a DNS.FAQ file which described the
round robin and multiple CNAMES described below...
I have been testing HP's implemenation of DNS and BIND 4.9.2 on
a Sparc Nameserver and there appears to be a major difference
from the expected results....
With the HP version I have to clear the Nameservers cache
(sig_named kill ; /etc/named ) and then query the desired site
each time to get the round-robin to work..(ie. receive
different ordering of IP addresses).
With the BIND 4.9.2 implementation each time I submit a
query...without the server caching the response...I get a
different ordering of the IP addresses!!!
Are these two scenarios both round-robin?
Any comments are most appreciated.
Cheers,
JP
>Though I have used and recommended multiple CNAME RR's in an >"RRset" (which it really isn't) in the past, I hereby >apologize and=