On Mon, Sep 1, 2008 at 3:43 AM, Marc Perkel <[EMAIL PROTECTED]> wrote:
>
>
> Aaron Wolfe wrote:
>>
>> On Sun, Aug 31, 2008 at 10:59 PM, RobertH <[EMAIL PROTECTED]> wrote:
>>
>
>>> It was explained somewhere earlier in the thread that he sometimes has
>>> to reboot his central dns servers and he apparently doesn't run local
>>> caching servers on the individual MX/SA nodes.
>>>
>>> I have to say (as others have mentioned in this thread and elsewhere)
>>> that running a local caching nameserver on any busy MX or SA server
>>> seems to solve this issue quite well without needing any scripts.  If
>>> you are rsyncing any zones from zen, etc. having the zone served up
>>> locally is awesome for quick lookups too.
>>>
>>> -Aaron
>>>
>>>
>
> Maybe my situation is unique. I'm running about 35 virtual servers and
> rather than run named in each one I have 3 virtual servers dedicated to
> doing caching DNS. One main one with 4 gigs allocated so that it caches for
> all of them and 2 backups in case something happens to the main one.
>
>
>

I think it's just a memory use vs. performance thing..  running a
nameserver in each instance might give you better performance and
stability, but of course it will use more ram.  Really though I don't
think named in a caching configuration is too bad of a pig on ram, and
there are high performance/low ram alternatives that just do caching.
I have a caching name server running on my home router (a linksys
thing that runs linux) and it has only 16MB ram total for the whole
system.  The nameserver used there is called dnsmasq and it appears to
use about 51`2k of ram in caching only mode.  Might be something to
consider since even with your script running every minute, you still
have (up to 60 sec)+(the time the script takes to run)+(the time it
takes everything to use the new resolv.conf) seconds of downtime that
could be avoided.

Reply via email to