Joining a new node to the bryce cluster

This procedure is usually done when we return an broken node to Angstrom and they send us back a new one. It's a bit simplistic, and prone to failure.

• Starting the node
• Re-imaging the node
  • Pick a source node
  • Configuring the network
  • Imaging with dd
• Configuring the network
• Generating new ssh keys
• Setting up automatic login

Starting the node

When you plug in a new node, it should just power on. If not, please read Powering on the bryce cluster nodes. The Titan64 nodes are preloaded with SUSE Linux Enterprise 8.0, and have a default root password of foobar

You will to bring the new node to single user mode, via shutdown now

Re-imaging the node

Our nodes are currently running SUSE 9.0, so the first thing we will do is re-image the node.

Because we have so few nodes, this procedure has not been exactified just yet. In the future, we would like to move to using the YaST autoinstall package.

Pick a source node

We will be imaging off another node. You can pick any node you like, but it's recommended that there is no user activity. If possible, you should shut the machine down to single user mode via shutdown now.

Once you have done this, take a network cable (I use blue ones to avoid confusion), and run it from the left (or upper) ethernet port on the source node, and plug it into the right (or lower) ethernet port on the new destination node.

Configuring the network

On the new (destination) node, run the following as root:

ifconfig eth0 192.168.25.1 netmask 255.255.255.0

On the source node, run the following as root:

ifconfig eth1 192.168.25.2 netmask 255.255.255.0

Make sure you can ping the new node from the source node:

ping 192.168.25.1

Imaging with dd

On the destination node, you will want to start a netcat server that pipes all of it's output directly to the hard disk device. Run this as root

netcat -l -p 5001 | dd of=/dev/hda

On the source node, you will want to do the opposite:

dd if=/dev/hda | netcat 192.168.25.1 5001

Once completed, you will want to unplug the ethernet cables and reboot the destination node.

Configuring the network

Plug a pink ethernet cable into the right-most ethernet port, and start the machine up. Run /sbin/YaST to reconfigure the IP address and hostname. I recommend rebooting after this.

Generating new ssh keys

Out of the box, your new machine will have the same ssh keys as your source node. To regenerate them, I recommend running this as root:

rm /etc/ssh/*key*
/etc/init.d/sshd stop
/etc/init.d/sshd start

Setting up automatic login

Copy the key data from /etc/ssh/ssh_host_rsa_key.pub on the new node. Paste it into /etc/ssh/known_hosts on bryce-n1, and then run clsync to sync that change to the other nodes.