Cloning a Slave Node

To clone a slave node, you will follow different procedures, depending upon whether or not the slave has a virgin or corrupted disk, or if the slave node has a running and operating version of linux on it.

Virgin disk / Not able to boot linux from own hard disk

Put a boot floppy disk in the machine
Verify that no other machines are currently being cloned (the master can't yet support parallel cloning!)
Verify that the machine is connected to the ethernet
Plug the keyboard and monitor cables for the second position of the console switch into the machine
If the machine is not turned on, turn it on now
Push the "soft reset' button on the front panel
When the machine is finished testing memory, you may with to do some or all of the following from firmware:

Update AlphaBIOS
Format the hard disk (it doesn't matter how you do this, but it is a good way to test disk & controller)

Now select the LinuxNFS boot option with the scrolling keys, and <RETURN>
The cloning procedure should take about ten minutes.
As soon as the machine begins to copy files (many will stream by!) take out the floppy disk. This is important: if you fail to do this the machine will reboot itself and begin the entire cloning procedure again and again!
When it is finished cloning, reboot the machine with the "soft reset" button. It should come up with a login: prompt.

System operating and able to boot from it's own hard disk

Verify that no other machines are currently being cloned (the master can't yet support parallel cloning!)
As root, rlogin to the machine from the master
run the script /home/install/auto-clone as follows. To re-clone node n017, type

/home/install/auto-clone n017

For a description of how this works, please see here

After a few seconds, your rlogin shell will exit (the node is no longer reachable on the network, at least, not by you!)
The cloning procedure is entirely automatic. If you monitor the activity on the master, you will notice a lot of work by rpc.nfsd (you can also see this from the graphical load monitor on the X display. After about five minutes the load should drop. After about another five minutes:
You should be able to rlogin to the node again.