DNS problems this morning -
Bob Goodwin - Zuni, Virginia, USA
bobgoodwin at wildblue.net
Tue Nov 13 19:12:27 UTC 2012
On 13/11/12 13:06, Rick Stevens wrote:
> You don't. You put the entries in the /etc/resolv.conf file and the
> resolver library picks them up.
>
>> The 5 second
>> timeout seems much to long when combined with 5 tries, perhaps fewer
>> tries would be better? However I imagine there were good reasons for
>> the defaulsts ...
>
> If you've ever run a big network (or a really popular one) you can watch
> the DNS servers get pummeled--especially if you have short TTLs set on
> the records. That being said, even a busy name server should respond in
> 5 seconds or less, so that seems reasonable.
>
> The default retry count is 2 (not 5) so the defaults as stated would
> result in a 10 second delay before the second DNS server is consulted.
> Yes, that seems an eternity, but not everyone has fast Internet access.
> There are still people with dial-up service (hard to believe, but
> they're out there). The standards were set up to accommodate these
> older environments. If you want a true giggle, look up RFC 1149,
> "Transmission of IP Datagrams on Avian Carriers" and be glad that it
> never caught on. :-)
>
> You can put in as long a timeout or as many retries as you want, but
> the library will limit timeouts to no more than 30 seconds (even if you
> specify 45) and no more than 5 retries (even if you specify 10). That's
> what the "silently capped" bit means.
I've tried the following:
# Generated by NetworkManager
nameserver 127.0.0.1
nameserver 192.168.1.1
nameserver 184.63.128.68
timeout:1
attempts:1
I moved 127.0.0.1 to the first line and added the last
two limitations.
The only way I have to judge time is watching the bottom of the
Firefox display where it tells me it's "Looking up" an address and
doing a number of reloads on a complex page, e.g.
http://www.weather.com/weather/tenday/23898. It appears to moving
through rapidly, I don't see it dwelling on "Looking up" but for a
fraction of a second, spending more time transferring data.
Is there a better way to test?
--
http://www.qrz.com/db/W2BOD
box7
More information about the users
mailing list