Diskless Linux Cluster How-To.

摘要

Diskless linux clustering is not yet a turn-key solution. The process of configuring a cluster of diskless linux machines requires many modifications to the stock linux operating system before they can boot cleanly. This guide will help the experienced linux user to take a set of Pre-eXecution Environment capable machines and configure them appropriately. The following is a guide designed for instructing a skilled administrator on the steps necessary for configuring a diskless linux cluster. It is assumed that the reader is familiar with TFTP, DHCP, Pre-eXecution boot Environment (PXE), configuring and compiling kernels, and has general unix knowledge. Be prepared for some trial and error. During the writing of this report, I rebooted a diskless node well over 100 times in order to perfect the configuration. With the aid of this report one should only have to reboot their cluster a few times until the configuration is in good working order. As an overview, one will begin by configuring a 'master' node from where the slave nodes will retrieve their data. First, the process will begin with the configuration of TFTP and DHCP, which will allow the slaves to get an IP, hostname, kernel, and ramdisk. Next, a file from the SysLinux UNIX tape archive is used to permit PXE booting. Later, a custom kernel and ramdisk are created exclusively for the slave nodes for their boot process. Finally, the slaves are configured to operate in a diskless environment. The Post- Install section provides insight into some additional useful configurations as well as how a few of the problems that were encountered along the way were resolved.