That is the idea exactly. All of the client nodes share a single
/etc/fstab.
So if I understand you correctly, if my /etc/fstab looks like the one
below the clients will ignore the "192.168.0.1:/opt/ltsp/i386" part and
get the NFS server from the DHCP response.
#bogus nfs server for client roots to make initrd work
192.168.0.1:/opt/ltsp/i386 / nfs dhcp 0 0
proc /proc proc
defaults 0 0
LTSP has init scripts that create a ramfs that the client uses to put
its random writable and device custom bits in.
Jeffrey Law wrote:
On Thu, 2007-10-25 at 12:44 -0600, David A. Kennel wrote:
Here are the constraints: I'm creating a
deployable package and I do not know the IP address or path of the
server that the client will need to connect to (the D/U version figures
that info out from the DHCP root-path option). If I include the dhcp
option in the /etc/fstab does it ignore the configured server and path
in the fstab and use what it receives from DHCP or does the correct
server need to be in the /etc/fstab?
IIRC, you don't need the correct information for the root filesystem
in the fstab. Ideally in the kind of environment you're describing
you want /etc/fstab to be shared if at all possible.
Presumably you're using tmpfs for /var and /tmp. How are you dealing
with the random bits in /etc that need to be writable?
Jeff
--
--
David Kennel
Los Alamos National Laboratory
CTN-1
_______________________________________________
Stateless-list mailing list
[email protected]
http://www.redhat.com/mailman/listinfo/stateless-list
