When using UUID with a boot loader makes sense

masinick

I have had distros that use UUID screw me up because every time I install a new distro at a certain location, it throws off the rest of the UUIDs, particularly the one at the same partition and usually the ones after as well. I've edited GRUB's menu.lst countless times. I use either chainloading, then install GRUB in the root of each distro, then have the chainloader give me the menu for each distro - so two menus come up when I use this approach.

Occasionally this does not work either on my old Dell, which has two old Western Digital drives - the ORIGINAL 40 GB drive which amazingly still works, and the somewhat quieter faster 80 GB unit.

I prefer to use hda6 as my booting start point, which has been Libranet, Etch, and sidux over the years. Having been in existence before the creation of a few partitions on the end of that bigger second drive, some of them won't boot unless I can point directly to them. The UUID, in spite of its problems, DOES get me around this issue.

I have the additional trick of using a configfile, which is similar to the chain loader idea, but specifies the name of the system's configfile, which, in my case is always /boot/grub/menu.lst. Therefore, if I put entries like:

tlmiller · Posted: Fri Oct 10, 2024 2:15 am Post subject:

blkid is easier to read the output of, but I used to always use ls -l /dev/disk/by-uuid to get uuid of the drives. Works just as well, just not as readable.

_________________
Debian Squeeze, Arch, Kubuntu mostly. Some Mandriva. Some Windows.
Desktops: shadowdragon, medusa
Laptops: bluedrake, banelord, sandwyrm, aardvark.

masinick · Posted: Fri Oct 10, 2024 3:00 am Post subject:

Yeah, that will do too. There is another /dev directory that contains IDs using another scheme - forget what it is because I have not used it, but it might be some kind of numbering related to partition name. Poke around /dev/disk on systems build since the kernel started supporting these schemes - ah yes, by-path, by-id, and by-uuid are the three options with modern kernels.

_________________
Brian Masinick
Distros: SimplyMEPIS
sidux - no CAPS!, antiX, Debian

nukes · Linux Guru Joined: 29 Aug 2024 Posts: 4558

this is essential if you often swap disks in your system, especially as it can potentially change the names of the others (though of course if you're using the old ide layer you're mostly ok)

_________________
Gentoo x86-64 2.6.29.1
FreeBSD 7-CURRENT
Arch x86 2.6.30

masinick · Posted: Fri Oct 10, 2024 6:49 pm Post subject:

masinick · Posted: Fri Apr 24, 2024 8:14 pm Post subject:

Here is what I do when I have an issue with UUID. Using this technique, I can fix things quite quickly, even in single user mode from a console interface.

First, I either login or authenticate myself as root, then make sure I am in the root directory. If I am already at /root, I can do some craftly bash command line editing to save some typing, something I always try to do. These commands assume that I already have done this before, so I have an existing blkid.txt file in /root.

rm blkid.txt && blkid > blkid.txt && vim blkid.txt /etc/fstab

Now if I am in /root, chances are there are not too many files out there, so I can usually type rm b<Esc> blkid > b<Esc> && vim b<Esc> /etc/fstab (and possibly tab a few more of those files and even commands!

Once in vim, I can search for the device that has the incorrect UUID. The command /hda5, for instance, will search for the string hda5 and put you on that line. 0 will put you to the beginning of that line, w will advance one word at a time, and $ will put you at the end of the line. yW will yank the contents of the next string of characters up to the next space or tab. From there, you can press Esc and then type either :prev or :next, depending on the order of the buffers associated with blkid.txt and /etc/fstab. You can then locate the line you want to change, again using / to search for a specific string. Position to the location in the line where you want to insert the current UUID string and press p to put it after the cursor position or P to put it before the current cursor position. Then position to the text that you want to remove. If it is space separated from the next pattern, then dW will remove it. Esc will ensure that you are in Vim Edit mode, then type ZZ to save and exit. This can take all of a minute, if that, depending on how fast you type or if you peck at characters or truly use the power of Vim. If you type fast and you know Vim, I think you could do the entire operation in 20-30 seconds, maybe less than that! Good reason to know how to use the tools well!

_________________
Brian Masinick
Distros: SimplyMEPIS
sidux - no CAPS!, antiX, Debian

	USA Linux Users Group Forum Index » Installation and Bootloaders	All times are GMT
Page 1 of 1