DNS Cache, database backends...

Forrest Aldrich forrie at navipath.com
Wed Feb 7 20:54:02 UTC 2001


I can understand some of the caveats with this.

However, my presumption was that someone "clueful" would be potentially 
using this optional feature.

In our case, we have 10k rpm disks, 2gig of RAM, and sync'd with the US 
Navy time service.  I think that's sufficient :)

More specifically, where you're servicing a huge array of dialpools (that 
use a given server as a forwarder), the ability to prime the server with a 
cache would be invaluable.

There are often legitimate reasons that require the server to be rebooted 
-- mostly to do with kernel updates and other facets of the OS that require 
rebooting.   Just a part of the work.

Sounds like others would be interested in an optional feature like 
this.   If it's created, it should be noted as optional along with all the 
caveats so that you can't BLAME someone for your stupidity :)


_F

At 04:34 PM 2/7/2001 +0000, James Raftery wrote:

>On Wed, Feb 07, 2001 at 10:55:57AM +0000, Jim Reid wrote:
> > Some computers are bad at timekeeping.
>
>Sure, but some computers are bad at disk I/O, some bad at shuffling
>around large amounts of data in real memory. They still use software
>that calls upon them to do these things.
>
> > I can envisage clueless DNS admins copying these
> > files from server to server and then whining because the restored
> > caches are not the same because the server's clocks don't keep the
> > same time.
>
>Yep, I could see that too. The answer to them is simple; tough. If they
>want time-based operations to work correctly they need to give their
>systems a cohesive view of time.
>
> > And to repeat what I said earlier, the easy way to solve
> > this "problem" is simply to leave the name servers alone and only stop
> > and start them when it's really necessary: for example when installing
> > new name server software.
>
>I agree. However I also think that the ability to prime ones cache upon
>startup either from a dump of some preexisting cache or from a list of
>desired cache contents I make up myself would be a useful feature. Even
>with the best administration nameservers need to be stopped from
>time-to-time. Not having to lose the contents of the cache would be
>neat. It's not essential, by any means, but I think it would be kinda
>groovy.
>
> > This discussion about preserving caches is just silly. It's a bit like
> > expecting the entire contents of your computer's RAM or the state of
> > every CPU register to be preserved after the system has been power
> > cycled. Sure it could be done, but what's the point? Really.
>
>My laptop can "suspend to disk" :-)
>
>
>ATB,
>
>james
>--
>James Raftery (JBR54)
>   "It's somewhere in the Red Hat district"  --  A network engineer's
>    freudian slip when talking about Amsterdam's nightlife at RIPE 38.



More information about the bind-users mailing list