We have a pxe-bootable live system to examine a node without touching the system on the harddrive. It is basically a self-made chroot environment.

usage

• start the script live.sh IP_address available in the git repository
  git+ssh://user@n0:srv/git/shared/cluster/scripts.git
  It creates the right pxeboot configuration for the nodes (see pxe configuration)
• reboot of the node
• use serial over LAN (SOL) using the IPMI card
  ipmitool -I lanplus -U user -P password -H IPMI_IP sol activate
  
• wait until you get a prompt

other access

One could also enter the system using rsh, ssh (right now only from a node which has not the common known_hosts) and via local access.

change or update the live system

The root-directory is a NFS mount to the FAIserver in /srv/live. Chroot into this directory:
chroot /srv/live /bin/bash
Do necessary updates via aptitude or similar ways. The live system is a lenny installation.

pxe configuration

The configuration is generated by the live.sh script. The pxe configuration looks like this:

default live
label live

kernel kernel/vmlinuz-2.6.23.1-faiinstall
serial 1,19200n8
append  root=/dev/nfs nfsroot=/srv/live,v3,tcp,rsize=32768,wsize=32768 ip=:::::eth1:dhcp console=ttyS1,19200n8

Since we want to use SOL the configuration of the serial entry is essential.

-- HenningFehrmann - 28 Aug 2008