>>>>> On Tue, 28 Feb 2006 08:05:27 +0100 (CET), 
>>>>> 6bone@6bone.informatik.uni-leipzig.de said:

> In my case, the bind resolvs some hundred thousand requests per day. After 
> nearly two days, the bind stops the resolving (no crash). The logfile 
> shows the following:

> Feb 28 05:08:03 gate named[2859]: /usr/src/dist/bind/lib/dns/cache.c:612: unexpected error:

(snip)

> The size of the named starts at nearly 6mb and is increasing to 99mb. If 
> the size smaller than 99mb, all works fine, at a size of 99mb the error 
> occurs.

> options {
>          directory "/etc/namedb";
>          listen-on-v6 { any; };
>          forwarders { aa.bb.cc.dd; ee.ff.gg.hh; };
>          notify no;
>          datasize 96M;
>          max-cache-size 64M;
> ...


> exists there any patches against the problem?

First off, limiting the datasize is typically not a good idea, unless
you have a very special reason for that.  I'd suggest remove it from
named.conf.  (Note that it is still necessary and effective to set
max-cache-size to a reasonable value.)

Secondly, one very easy but pretty effective remedy for this type of
problems is to use internal memory allocator, especially when you are
not running named on a multi-processor box with multiple threads.  You
can do that by rebuilding BIND9 with ISC_MEM_USE_INTERNAL_MALLOC being
non-zero.

And finally, this looks like a general BIND9 issue rather than a
NetBSD-specific topic.  I'd suggest you ask this type of question at,
e.g., bind-users@isc.org.

					JINMEI, Tatuya
					Communication Platform Lab.
					Corporate R&D Center, Toshiba Corp.
					jinmei@isl.rdc.toshiba.co.jp