Help: BIND 8.2.2_P5 hangs on Red Hat 6.2

[Mark.Andrews at nominum.com]

	There was a problem with syslog and BIND whereby if syslogd
	locked up BIND would stop under Linux.  Ensure that you have
	the latest syslog code, incl. syslogd.

	There are also cases where named and syslog get into a deadly
	embrace where syslog wants to resolve a IP address and named
	wants to log messages.

	Mark

> 
> I know this problem is not unique to us because I found some
> discussion of it in the Dejanews archives.  What I did not find was a
> definitive resolution.
> 
> One of our secondary name servers keeps entering a state where it
> stops serving requests.  Sometimes we go for weeks without seeing the
> problem, and sometimes it happens multiple times in a single day.  We
> have no idea what triggers it.
> 
> This is a Red Hat 6.2 system with all current updates (including
> sysklogd) and Linux kernel 2.2.17.
> 
> When it is in this state, it syslogs large numbers of messages like this:
> 
>   named[567]: ns_req: sendto([192.168.1.1].4887): Connection refused
> 
> The IP addresses and port numbers vary wildly, but the messages are
> plentiful; named just seems unable to send responses at all.  (Or
> maybe it is sending them too late and the other end is giving ICMP
> error responses.)
> 
> This is the only system on which we have observed the problem.  It is
> unique for us in two ways:
> 
>   - It is the only secondary we have which is not a "stealth"
>     secondary, so it is the only one receiving NOTIFYs.
> 
>   - It is our "loghost", meaning it sees an unusual amount of UDP
>     syslog activity.
> 
> Any suggestions would be much appreciated.  Thanks!
> 
>  - Pat
> 
> 
--
Mark Andrews, Nominum Inc.
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark.Andrews at nominum.com