Re: running amdump manually at the moment with extra debugging enabled

On 28/09/2007, Jean-Louis Martineau <martineau AT zmanda DOT com> wrote:

Craig Dewick wrote:
>
> 6 root@jedi #----> more dumper.20070928130458000.debug
> dumper: debug 1 pid 15243 ruid 982 euid 0: start at Fri Sep 28
> 13:04:58 2007
> dumper: debug 1 pid 15243 ruid 982 euid 0: rename at Fri Sep 28
> 13:04:58 2007
> dumper: time 173.205: connect_port: Try port 0: Available -
> dumper: time 173.212: connect_portrange: connect from ::.35127 failed:
> Network is unreachable
> dumper: time 173.213: connect_portrange: connect to ::1.35126 failed:
> Network is unreachable
> dumper: time 173.213: stream_client: Could not bind to port in range 0-0.
localhost resolv to '::1', which is an IPv6 address, but you doesn't
have IPv6 configured.

Why localhost resolv to '::1' instead of 127.0.0.1?
Check your host table configuration, starting from /etc/nsswitch.conf

amanda use the getaddrinfo function to get the address of localhost.

Ok I found this:

14 craig@jedi #----> more /etc/inet/ipnodes
#
# Internet host table
#
::1             localhost
127.0.0.1       localhost

which might explain the problem! I found a reference to the ipnodes file on the man page for getaddrinfo:

FILES
     /etc/inet/hosts
           host name database

     /etc/inet/ipnodes
           local database that associates names of nodes with IP
           addresses

     /etc/netconfig
           network configuration database

     /etc/nsswitch.conf
           configuration file for the name service switch

and I'd never given any thought to the ipnodes file before. 8-) I have commented out the lines referring to "::1       localhost" in /etc/inet/ipnodes on all three of the Solaris systems and will see what happens in just over two hours when the next cron-scheduled run of 'amdump' occurs.

I don't use IPv6 on any of my systems, but it's possibly a hangover from originally installing Solaris as there's an option to 'enable' IPv6 when installing Solaris. I don't have an /etc/hostname6.hme0 on the tape server host though it could have had that at some stage in the past. Maybe partial IPv6 support stays enabled in the kernel even when the ipv6 version of the /etc/hostname file doesn't exist...

I'll let you know if the new discovery solves the problems that have been occuring.

Craig.

--
Craig Dewick ( craig AT lios.apana.org DOT au). http://lios.apana.org.au/~craig
RailZone Australia: http://www.railzone.org ( craigd AT railzone DOT org)
The SunShack: http://www.sunshack.org ( cdewick AT sunshack DOT org)
ClassicSaab: http://www.classicsaab.net ( c900 AT classicsaab DOT net)