DragonFly users List (threaded) for 2004-12
Re: looking for diskless advice
:> If you have normal machines e.g. with hard disk and all, I would
:> use a small alternative root filesystem, which automatically fetches or
:> rsyncs the real root + usr filesystem. I'd configure loader or
:> boot0 to switch the default config on demand and restart.
:Yup, these are normal machines. What is an example of an "alternative root"?
That would be like e.g. creating two slices and putting a mini-system
on one slice and the real system on another slice.
I personally think that's overkill. If the boxes have CD drives then
it is just as easy to stick a live CD in if a machine becomes unbootable.
For normal machines I would recommend using diskless booting to automate
or semi-automate the initial install, but to then have a normal
installed system running on each box with scripts to keep them up
to date. For updates, one master machine with the latest built world
on it can export /usr/src and /usr/obj via NFS and the cluster of boxes
can be updated basically just with a make installworld.
Note that you can setup a server to export the ISO release image with
NFS, and machines diskless-booted with that will operate just as if you
had booted from CD. What I would do is cpdup an ISO into a directory
on your NFS server, set up a DHCP server if you don't already have one,
and then modify the image so diskless-booted machines come up with a
working network, sshd, and key so you can access them remotely from the
server without any manual work on their consoles. Then automate the
installation using ssh/scp/scripts from the server.