PXE Magic: Flexible Network Booting with Menus
Now that TFTP is set up, all that is left to do is to install the syslinux package (available for most distributions, or you can follow the installation instructions from the project's main Web page), copy the supplied pxelinux.0 file to /var/lib/tftpboot (or your TFTP directory), and then create a /var/lib/tftpboot/pxelinux.cfg directory to hold pxelinux configuration files.
You can configure pxelinux with or without menus, and many administrators use pxelinux without them. There are compelling reasons to use pxelinux menus, which I discuss below, but first, here's how some pxelinux setups are configured.
When many people configure pxelinux, they create configuration files for a machine or class of machines based on the fact that when pxelinux loads it searches the pxelinux.cfg directory on the TFTP server for configuration files in the following order:
Files named 01-MACADDRESS with hyphens in between each hex pair. So, for a server with a MAC address of 88:99:AA:BB:CC:DD, a configuration file that would target only that machine would be named 01-88-99-aa-bb-cc-dd (and I've noticed it does matter that it is lowercase).
Files named after the host's IP address in hex. Here, pxelinux will drop a digit from the end of the hex IP and try again as each file search fails. This is often used when an administrator buys a lot of the same brand of machine, which often will have very similar MAC addresses. The administrator then can configure DHCP to assign a certain IP range to those MAC addresses. Then, a boot option can be applied to all of that group.
Finally, if no specific files can be found, pxelinux will look for a file named default and use it.
One nice feature of pxelinux is that it uses the same syntax as syslinux, so porting over a configuration from a CD, for instance, can start with the syslinux options and follow with your custom network options. Here is an example configuration for an old CentOS 3.6 kickstart:
default linux label linux kernel vmlinuz-centos-3.6 append text nofb load_ramdisk=1 initrd=initrd-centos-3.6.img ↪network ks=http://10.0.0.1/kickstart/centos3.cfg
The standard sort of pxelinux setup works fine, and many administrators use it, but one of the annoying aspects of it is that even if you know you want to install, say, CentOS 3.6 on a server, you first have to get the MAC address. So, you either go to the machine and find a sticker that lists the MAC address, boot the machine into the BIOS to read the MAC, or let it get a lease on the network. Then, you need to create either a custom configuration file for that host's MAC or make sure its MAC is part of a group you already have configured. Depending on your infrastructure, this step can add substantial time to each server. Even if you buy servers in batches and group in IP ranges, what happens if you want to install a different OS on one of the servers? You then have to go through the additional work of tracking down the MAC to set up an exclusion.
With pxelinux menus, I can preconfigure any of the different network boot scenarios I need and assign a number to them. Then, when a machine boots, I get an ASCII menu I can customize that lists all of these options and their number. Then, I can select the option I want, press Enter, and the install is off and running. Beyond that, now I have the option of adding non-kickstart images and can make them available to all of my servers, not just certain groups. With this feature, you can make rescue tools like Knoppix and memtest86+ available to any machine on the network that can PXE boot. You even can set a timeout, like with boot CDs, that will select a default option. I use this to select my standard Knoppix rescue mode after 30 seconds.
Because pxelinux shares the syntax of syslinux, if you have any CDs that have fancy syslinux menus, you can refer to them for examples. Because you want to make this available to all hosts, move any more specific configuration files out of pxelinux.cfg, and create a file named default. When the pxelinux program fails to find any more specific files, it then will load this configuration. Here is a sample menu configuration with two options: the first boots Knoppix over the network, and the second boots a CentOS 4.5 kickstart:
default 1 timeout 300 prompt 1 display f1.msg F1 f1.msg F2 f2.msg label 1 kernel vmlinuz-knx5.1.1 append secure nfsdir=10.0.0.1:/mnt/knoppix/5.1.1 ↪nodhcp lang=us ramdisk_size=100000 init=/etc/init ↪2 apm=power-off nomce vga=normal ↪initrd=miniroot-knx5.1.1.gz quiet BOOT_IMAGE=knoppix label 2 kernel vmlinuz-centos-4.5-64 append text nofb ksdevice=eth0 load_ramdisk=1 ↪initrd=initrd-centos-4.5-64.img network ↪ks=http://10.0.0.1/kickstart/centos4-64.cfg
Each of these options is documented in the syslinux man page, but I highlight a few here. The default option sets which label to boot when the timeout expires. The timeout is in tenths of a second, so in this example, the timeout is 30 seconds, after which it will boot using the options set under label 1. The display option lists a message if there are any to display by default, so if you want to display a fancy menu for these two options, you could create a file called f1.msg in /var/lib/tftpboot/ that contains something like:
----| Boot Options |----- | | | 1. Knoppix 5.1.1 | | 2. CentOS 4.5 64 bit | | | ------------------------- <F1> Main | <F2> Help Default image will boot in 30 seconds...
Notice that I listed F1 and F2 in the menu. You can create multiple files that will be output to the screen when the user presses the function keys. This can be useful if you have more menu options than can fit on a single screen, or if you want to provide extra documentation at boot time (this is handy if you are like me and create custom boot arguments for your kickstart servers). In this example, I could create a /var/lib/tftpboot/f2.msg file and add a short help file.
Although this menu is rather basic, check out the syslinux configuration file and project page for examples of how to jazz it up with color and even custom graphics.
Kyle Rankin is a systems architect; and the author of DevOps Troubleshooting, The Official Ubuntu Server Book, Knoppix Hacks, Knoppix Pocket Reference, Linux Multimedia Hacks, and Ubuntu Hacks.
Free Webinar: Hadoop
How to Build an Optimal Hadoop Cluster to Store and Maintain Unlimited Amounts of Data Using Microservers
Realizing the promise of Apache® Hadoop® requires the effective deployment of compute, memory, storage and networking to achieve optimal results. With its flexibility and multitude of options, it is easy to over or under provision the server infrastructure, resulting in poor performance and high TCO. Join us for an in depth, technical discussion with industry experts from leading Hadoop and server companies who will provide insights into the key considerations for designing and deploying an optimal Hadoop cluster.
Some of key questions to be discussed are:
- What is the “typical” Hadoop cluster and what should be installed on the different machine types?
- Why should you consider the typical workload patterns when making your hardware decisions?
- Are all microservers created equal for Hadoop deployments?
- How do I plan for expansion if I require more compute, memory, storage or networking?
|Speed Up Your Web Site with Varnish||Jun 19, 2013|
|Non-Linux FOSS: libnotify, OS X Style||Jun 18, 2013|
|Containers—Not Virtual Machines—Are the Future Cloud||Jun 17, 2013|
|Lock-Free Multi-Producer Multi-Consumer Queue on Ring Buffer||Jun 12, 2013|
|Weechat, Irssi's Little Brother||Jun 11, 2013|
|One Tail Just Isn't Enough||Jun 07, 2013|
- Speed Up Your Web Site with Varnish
- Containers—Not Virtual Machines—Are the Future Cloud
- Linux Systems Administrator
- Lock-Free Multi-Producer Multi-Consumer Queue on Ring Buffer
- Senior Perl Developer
- Technical Support Rep
- Non-Linux FOSS: libnotify, OS X Style
- UX Designer
- RSS Feeds
- It is quiet helping
56 min 32 sec ago
1 hour 13 min ago
- Reachli - Amplifying your
2 hours 29 min ago
3 hours 18 min ago
- good point!
3 hours 21 min ago
- Varnish works!
3 hours 30 min ago
- Reply to comment | Linux Journal
4 hours 22 sec ago
- Reply to comment | Linux Journal
6 hours 26 min ago
- Reply to comment | Linux Journal
10 hours 26 min ago
- Yeah, user namespaces are
11 hours 42 min ago