bind9 stops responding to ipv6 requests after a period

[Jakub Suchy]

Hi,
i am administrating bind9-9.2.4-0backports at our campus network with +3000
users.
We managed to get IPv6 to work (smtp, dns, peering enabled), but i am
experiencing problems with IPv6 on bind9.

Everything is configured properly and works, but after a period (sometimes a
day, hour, week), bind stops responding to IPv6 request totally:

client:~# host slovakia 2001:718:2::6
;; connection timed out; no servers could be reached

when i look to logs, i found nothing (am not sure, but i think i found nothing
also when enabled querylog), netstat shows:
tcp6       0      0 *:domain                *:*                     LISTEN
so bind should listen...

i restart bind /etc/init.d/bind9 restart and voila:
client:~# host slovakia 2001:718:2::6
Using domain server:
Name: 2001:718:2::6
Address: 2001:718:2::6#53
slovakia.sh.cvut.cz has address ...

it's VERY weird, does anybody have such problem? it seems as some bug in bind9
more configuration:
Intel XEON, 1GB RAM, kernel 2.4.27-1-686, libc6 2.2.5-11.5
listen-on-v6 port 53 { any; };
bind9 chrooted
it should NOT be internet connection problem, client -> server is 1 hop through
one cisco

Thanks,
Jakub Suchy
hostmaster at sh.cvut.cz

-- 
-- Jakub Suchy <jakub.suchy at logios.cz>
-- GPGkey: http://nemadema.cz/meba/gpg.key
-- LOGIOS NETWORK SYSTEMS
-- QUADRA s.r.o., Šaldova 217/7, 400 01 Ústí nad Labem
-- e-mail: info at logios.cz, web: http://www.logios.cz/