After verifying that all library versions are the same, the crash still occurs.
The backtrace is the same.  But note that line #2 probably indicates that the
stack was damaged.  However the log message is slightly longer, perhaps it
provides a clue.
Jul 16 21:14:03 moya named[19051]: rbtdb.c:1147: INSIST(noderefs != 0) failed
Jul 16 21:14:03 moya named[19051]: exiting (due to assertion failure)
Jul 16 21:14:03 moya named[19051]: rbtdb.c:1174: REQUIRE(prev > 0) failed
Jul 16 21:14:03 moya named[19051]: exiting (due to assertion failure)

JINMEI Tatuya / ?$B?@L@C#:H [Jinmei_Tatuya@isc.org] wrote:
> At Mon, 14 Jul 2008 01:08:51 -0700,
> Bob McElrath <bob_bind@mcelrath.org> wrote:
> 
> > With bind 9.4.2, I am seeing a regular assertion and segfault:
> >     Jul 13 22:26:00 moya named[27422]: rbtdb.c:1174: REQUIRE(prev > 0) failed
> >     Jul 13 22:26:00 moya named[27422]: exiting (due to assertion failure)
> > 
> > This is on debian, the package version is 9.4.2-10, on a DEC/Compaq alphaserver.
> > Bind runs for around 24 hours properly before it crashes.  The crash leaves a
> > core, so here is the backtrace:
> > 
> > #0  0x0000020000544c78 in raise () from /lib/libc.so.6.1
> > #1  0x0000020000546574 in abort () from /lib/libc.so.6.1
> > #2  0x0000000120020fc0 in ?? ()
> > #3  0x00000200000dab44 in dns_rbt_create () from /usr/lib/libdns.so.32
> > #4  0x00000200000e3e14 in dns_rbtdb_create () from /usr/lib/libdns.so.32
> > #5  0x0000020000095318 in dns_db_detachnode () from /usr/lib/libdns.so.32
> > #6  0x0000020000143058 in dns_resolver_shutdown () from /usr/lib/libdns.so.32
> > #7  0x0000020000147c74 in dns_resolver_prime () from /usr/lib/libdns.so.32
> > #8  0x000002000047363c in isc_taskmgr_create () from /usr/lib/libisc.so.32
> > #9  0x00000200004dfefc in start_thread () from /lib/libpthread.so.0
> > #10 0x00000200005f4734 in clone () from /lib/libc.so.6.1
> > 
> > Any suggestions on how to fix this?  My DNS setup is exceedingly simple, just a
> > couple domains (one of which is my email address if you want to interrogate
> > bind).
> 
> This backtrace doesn't make sense (the call sequence doesn't match the
> actual code, etc).  Are you sure you have the right pair of named
> binary and core?
> 
> ---
> JINMEI, Tatuya
> Internet Systems Consortium, Inc.
--
Cheers,
Bob McElrath [ CERN Theory Division ]

"The only way of discovering the limits of the possible is to venture a little
way past them into the impossible." -- Arthur C. Clarke, 1917-2008