Tom H wrote: > Bob Proulx wrote: > > Also I know that RHEL/CentOS at least also moved from an autoraid of > > 0xFD to an explicitly mounted system too for the same reasons. But > > they do it by specifying the UUIDs on the kernel command line from > > grub. It makes for some very long command lines. I like the Debian > > choice better. > > Dracut, the initramfs infrastructure that RH uses and that I often use > on Debian, doesn't require the UUID of an array on the kernel command > line.
It isn't a dracut requirement. But it is how a default installation
is set up in the grub.conf file. Specifically I am talking about 6.2
where I needed to dig through it in detail to add an additional raid
array. Try a pristine installation using raid and you will see it set
up the rd_MD_UUID= lines. For example:
title CentOS (2.6.32-220.13.1.el6.x86_64)
root (hd0,0)
kernel /vmlinuz-2.6.32-220.13.1.el6.x86_64 ro root=/dev/mapper/f1-root
rd_LVM_LV=f1/root rd_NO_LUKS LANG=en_US.UTF-8 nomodeset
SYSFONT=latarcyrheb-sun16 crashkernel=auto
rd_MD_UUID=f94ed1ad:5e420791:0bf0cfe7:310c77d1 rd_LVM_LV=f1/swap
KEYBOARDTYPE=pc KEYTABLE=us rd_NO_DM
initrd /initramfs-2.6.32-220.13.1.el6.x86_64.img
In the above *only* that specific UUID raid array will be set up.
Adding a second array requires adding a second rd_MD_UUID entry. It
is somewhat insidious because lvm will try to work with only part of
the physical extents if it is possible. So adding an additional array
actually worked without this for about a month until I increased the
size of the root logical volume and only then did it fail to boot. I
discovered that I needed to add the additional UUID in order to have
it assembled. Or remove it and use autoraid.
But that is just a default. I converted my system to autoraid (by
removing the rd_MD_UUID entries) because it was simpler.
But I don't want to talk about that system here. I just mentioned it
to note that they also moved away from autoraid assembly. They just
do it in a different way.
Bob
signature.asc
Description: Digital signature
