NetBooting

From MDWiki
Jump to navigationJump to search

The cluster nodes are imaged and booted by a network boot process.

The mango nodes should be connected to the network switch by the second (ipmi) ethernet interface. That interface should be configured within the BIOS to network boot. (It is acceptable for this to be the second boot preference so that only unimaged disks proceed to net-boot)

The merlot nodes should be connected via their first interface.

If the node enters network-boot, it issues a DHCP broadcast request for an IP number.

The Lychee DHCP service recognises the node MAC address (within /etc/dhcpd.conf) and responds. This DHCP response includes "nextserver" (also lychee) and "file" parameters which the node PXE boot bios uses as a "tftp server" request to load and runs the CentOS "pxelinux" loader.

The "pxelinux" looks in the server "pxelinux.cfg" for a file that matches either the MAC address, IP address, IP range or a default.

This should load a configuration file such as "kickstart_x2200.cfg". As the name suggests, this file contains the command to run an OS kickstart installation.

The kickstart uses http to collects the kickstart file and the OS distribution. These are made available by the Lychee http server from directory /var/www/html/.

Upon completion of the OS installation the node should reboot and then boot the disk.