ndc problem

[Mark.Andrews at nominum.com]

> >>>>> "Jim" == Jim Tinlin <jt at jumbo.surfari.net> writes:
> 
> 
>     Jim> BIND starts and runs fine, however ndc doesn't.  When I first
>     Jim> start BIND, I get the following log entries:
> 
>     Jim> Jan 16 15:32:25 cachecow named[28906]: unix control "/var/run/ndc" 
> connect failed: Invalid argument 
>     Jim> Jan 16 15:32:25 cachecow named[28906]: ctl_server: setsockopt(REUSE
> ADDR): Operation not supported on transport endpoint
>     Jim> Jan 16 15:32:25 cachecow named[28906]: ctl_server: bind: Address al
> ready in use
> 
>     Jim> This even happens when I re-boot the server.  Now here's the
>     Jim> funny (?)  part...I took a look at the file that ndc creates
>     Jim> in /var/run:
> 
>     Jim> srw------- 1 root root 0 Nov 12 08:10 ndc=
> 
>     Jim> The date is the last date that I restarted BIND *before* I
>     Jim> updated it...  I compiled BIND on my server on 11/15...the
>     Jim> above date is 11/12!  Now this is probably the root of the
>     Jim> problem, and how would I go about getting rid of this file
> 
> rm should delete this file provided nothing has it open at the
> time. Even then, it should be deleted OK.
> 
> It looks like this file didn't get deleted by the previous version of
> BIND that you ran. This was a known bug for some OS platforms.

	It was a known bug in ealier releases of BIND 8.  These are
	addressed with BIND 8.2.2.  This is a *new* error EINVAL on
	connect.  The two expected errors on connect() are ECONNREFUSED
	and ECONNRESET.  Earlier versions only coped with one of
	these.

	Earlier versions produced spurious "ctl_server: bind: Address
	already in use" errors.  With the current version this is
	meaningful as the checks for old sockets are now done *before*
	the an attempt is made to create the control socket.

> 
>     Jim> (I've tried a number of things...nothing 'simple' seems to do
>     Jim> the trick)???  If I get rid of this, I assume ndc will create
>     Jim> another when it starts...which is why it's not starting
>     Jim> now...the ndc file already exists (but it thinks someone else
>     Jim> started it and won't continue).
> 
> This file is created by named, not ndc. ndc uses this socket to talk
> to the name server. If the file exists - say because named didn't exit
> properly - named gets very upset as you can see from the above error
> messages. So you need to delete this file and restart the name
> server.

	Just saying "delete it" does not address the underlying problem.

	Mark
--
Mark Andrews, Nominum Inc. / Internet Software Consortium
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark.Andrews at nominum.com