The only way I can think to pull this off is if you have a partition on that machine that you can hose completely. Keeping an ssh or telnet session alive requires too many binaries to be there, a shell, ssh, auth... too much has to be running for you to be able to overwrite anything "while" you're using it. Here's what I can think of... say you have a 300Mb partition, /dev/sdc1 for example, that was mounted as /home that you can hose without killing an ssh connection and that you can do without for the most part.
Take a machine right next to you, in your office, and install something slim onto it, a slackware 8.1 with sshd and a base filesystem is the best bet. Make sure its all on 1 partition. Make an image of that, just tar the whole thing up if you like... then scp that onto the remote machine. Then format over /dev/sdc1 and copy the image onto it. Then cd over to the fstab for the new slack partition and make certain that its fstab is correct for the hardware it is now on, for example:
/dev/sda1 swap swap defaults 0 0
/dev/sdc1 / ext3 defaults 1 1
Also change its /etc/rc.d/rc.inet2 for the IP address of the machine its now on...
Make certain its /etc/rc.d/rc.netdevice has in it the right modules for the new machine its on.
Might want to remember /etc/hosts too, but thats pretty perfunctory.
Now, make an entry in lilo or grub for the new slackware kernel and with /dev/sdc1 as the root filesystem. Then of course re-run lilo or grub. Reboot the machine, cross your fingers and hail a few penguins.
Now, take that machine right next to you with the slim-slack install on it, cut up its drives and make a mirror to the RedHat 7.3 install you want minus whatever you had in /home.
On the remote slack machine start formatting the old partitions.
Start copying the chunks over, partition by partition, or even just in clumps of tar files. Again, remember to make certain the /etc/fstab is right, the /etc/sysconfig/networking-scripts/ifcfg-eth0 script is alright, whatever the heck file it is for the network card module... probably /etc/modules.conf, and make certain to remove all of the /etc/rc.d/rcX.d and /etc/init.d entries for kudzo... actually any other cutsie redhat utility that may interrupt the boot sequence on the other side that doesn't have a timeout. Also... make certain you change its /etc/inittab entry to 3 as RH loves to default to 5 and who knows if X will hang the box....
Then of course when your done, use Slack's lilo to over-write the MBR again, this time with the default kernel pointed at the new RH partition, pray you set everything up right, and reboot.
That's the only way I can think to do it, in 2 steps. The upshot to it is that you can sandbox test it in your office with two lan boxes... the downshot is that this may suck time.
Last edited by finegan; 09-04-2002 at 09:20 PM.