disabling caching/growth of cache (was: Re: 4 interfaces, 4 different responses, 2 forwardings and 1 cache. How ?)

Barry Margolin barmar at bbnplanet.com
Mon Oct 25 04:22:49 UTC 1999


In article <Pine.LNX.4.10.9910231412000.14295-100000 at ann.ied.com>,
Jan Vicherek  <honza at ied.com> wrote:
>  couldn't I accomplish the effect of "no forward" ( no attempt to contact
>any DNS servers except itself ) by specifying pretty blank "." zone db
>file ?

Yes.

> Or would the DNS server, even if I have mostly blank "." zone db, attempt
>to contact some of the DNS servers that are contained in the NS records of
>the 10 zones it is serving ?

No.  Since it's authoritative for those zones, it should never need to
contact any other servers for those zones.

>
>> > Would the following configuration do it ? :
>> 
>> Except for the above options, yes.
>> 
>> Note that the eth0 named will maintain its own cache.  It will simply be a
>> subset of what's in the lo named's cache, since it will learn everything
>> (except its authoritative zones) by forwarding to the lo named.
>
>  so does that mean that there is no way to turn caching off ?

There's a compile-time macro that specifies the maximum cache time.  I
suppose you could set this to a really low number and recompile this
instance of named.  It will maintain a cache, but entries will be timed out
very quickly.

> Since I have to run multiple nameds to resolve queries to the outside, it
>would be nice if I didn't have to have two almost identical caches. But if
>that's not possible, that's OK, I've just upgraded to 64MB RAM, so I can
>live with it. Is there a way to say "the max. cache you can keep is 0kB or
>1kB ? I.e. if I specify option datasize XXX kB, where XXX is the amount
>right after start (taken from `ps auxwww | grep named`) plus 5, when the
>limit is reached, the named will : 
>
>A)not function correctly (I.e. refuse to answer queries fully & correctly)
>B)get rid of some cache to make sure that it can function normally.

C)the process will exit.  That option simply uses the system's setrlimit()
function to set the maximum size of the data segment, and when it can't get
any more memory it aborts.

-- 
Barry Margolin, barmar at bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.


More information about the bind-users mailing list