DHCP Help Again

Robert G. Brown rgb at phy.duke.edu
Thu Apr 4 06:52:47 PST 2002


On Wed, 3 Apr 2002, Adrian Garcia Garcia wrote:

For one thing don't use the range statement -- it tells dhcpd the range
of IP numbers to assign UNKNOWN ethernet numbers.  You are statically
assigning an IP number in your "free" range to a particular host with a
KNOWN ethernet number below.  I don't know what dhcpd would do in that
case -- something sensible one would hope but then, maybe not.  The
range statement is really there so you can dynamically allocate
addresses from the range to hosts you may never have seen before that
you don't care to ever address by name (as they might well get a
different IP number on the next boot).  

DHCP servers run by ISP's not infrequently use the range feature to
conserve IP numbers -- they only need enough to cover the greatest
number of connections they are likely to have at any one time, not one
IP number per host that might ever connect.  Departments might use it to
give IP numbers to laptops brought in by visitors (with the extra
benefit that they can assign a subnet block that isn't "trusted" by the
usual department servers and/or is firewalled from the outside by an
ip-forwarding/masquerading host).

You want "only" static IP's in your cluster, as you'd like nodo1 to be
the same machine and IP address every time.

Be a bit careful about your use of domain names.  As it happens, I don't
find cluster.org registered yet (amazingly enough!) but it is pretty
easy to pick one that does exist in nameservice in the outside world.
In that case you'll run a serious risk of routing or name resolution
problems depending on things like the search order you use in
/etc/nsswitch.conf.  Even my previous example of rgb.private.net is a
bit risky.

You should run a nameserver (cache only is fine) on your 192.168.1.1
server, presuming it lives on an external network and you care to
resolve global names.

Similarly you may want:

 option routers		192.168.1.1;

if you want internal hosts to be able to get out through your (presumed
gateway) server.

Finally, if you want nodo1 to come up knowing its own name without
hardwiring it in on the node itself, add

 option host-name	nodo1;

to its definition.

I admit that I do tend to lay out my dhcpd.conf a bit differently than
you have it below but I don't think that the differences are
particularly significant, and you have a copy of the one I use anyway if
you want to play with the pieces.  You should find a log trace of
dhcpd's activities in /var/log/messages, which should help with any
further debugging.

On your nodo1 host, make sure that:

cat /etc/sysconfig/network-scripts/ifcfg-eth0
DEVICE=eth0
BOOTPROTO=dhcp
ONBOOT=yes

and

cat /etc/sysconfig/network
NETWORKING=yes
HOSTNAME=nodo1

and that in /etc/modules.conf there is something like:

cat /etc/modules.conf
alias parport_lowlevel parport_pc
alias eth0 tulip

(or instead of tulip, whatever your network module is).

If you then boot your e.g. RH client it SHOULD just come up,
automatically try to start the network on device eth0 using dhcp as its
protocol for obtaining and IP number, ask the dhcp server for an address
and a route, and just "work" when they come back.

  Hope this helps.

       rgb

> server-name "server.cluster.org"
>  
> subnet 192.168.1.0 netmask 255.255.255.0
> {
>   range 192.168.1.2         192.168.1.10   #my client has the ip
> 192.168.1.2
>                                                                 #and my
> server the static ip 192.168.1.1
>  option subnet-mask                             255.255.255.0;
>  option broadcast-address                    192.168.1.255;
>  option domain-name-server                 192.168.1.1;  
>  option domain-name                            "cluster.org";
>  
>  host  nodo1.cluster.org
>  {
>     hardware ethernet 00:60:97:a1:ef:e0; #here is the address of the
> client's card
>     fixed-address        192.168.1.2;
>  }
> } 
>  
> And finally some files on my server.
>  
> NETWORK
> ------------------------------------------
> networking = yes
> hostname =server.cluster.org
> gatewaydev = eth0
> gatewaye=
> ------------------------------------------
>  
> HOSTS ( In my server and in the client I have the same on this file )
> ------------------------------------------
> 127.0.0.1             localhost
> 192.168.1.1         server.cluster.org
> 192.168.1.2         nodo1.cluster.org
>  
>  
> Ok thats the information, I am a little confuse, could you help me please
> =). I can´t detect the mistake, I dont know if is the server or some card
> =s. Thanks for all.
> 
> ________________________________________________________________________________
> Get your FREE download of MSN Explorer at http://explorer.msn.com.
> _______________________________________________ Beowulf mailing list,
> Beowulf at beowulf.org To change your subscription (digest mode or
> unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
> 

-- 
Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu






More information about the Beowulf mailing list