On Wed, 12 Oct 2005 13:12:37 +0100
Matt Doidge <[log in to unmask]> wrote:
> Heya,
>
> After an quick cost-benifit analysis I decided that doing a full
> reinstall of all our dcache nodes in one day could be construed as an
> act of extreme masochism, so I have decided to go down the saner root
> of simply reinstalling the admin node, keeping the same name for it,
> and then reinstalling the two pool nodes that look off.
>
> > How will you keep host certificates on the poor things then, if they
> > are named by DHCP? You see, Grid machines are like Golems, if you
> > can name them you can control them :-)
>
> We tell ISS the MAC addresses of our machines, and the names we would
> like to give them, and they work their mojo. Thus our machines are
> named by us, but in a slightly round about way.
>
> right, off i go to get physical with the machines,
> matt
Just so you know (In case you don't), you can take physically the
network card from a first computer and place it in the second and then
DHCP will think the second computer is the first computer, and assign
the host name and IP address of the first computer to the second
computer.
My big worry about all this is you need control of a DHCP server to
perform network booting which is a dependency for auto installation
which to me is one of the best cluster management technologies I could
think of.
Do you have a kick start server?
Regards
Owen
|