Hi Luv' ers I am having a grub install problem. I installed 13.10 and in the process destroyed grub and I can't boot anymore. I can however boot using supergrub2 cd. (..so no panic!) If I boot from hard drive, I get error: dev with UUID xxxx... can't be found. (..something to that effect) I know that, because UUID xxxx... is now UUID yyyy .... I changed all reference to xxxx... to yyyy... in all grub.cfg files. (this box has 3 versions of Ubuntu) When I try to reinstall grub with grub-install /dev/sdb (note: boot HD is second one, but even if I select sda, I get same error) I get errors as follows: /usr/sbin/grub-bios-setup: warning: your core.img is unusually large. It won't fit in the embedding area. /usr/sbin/grub-bios-setup: error: embedding is not possible, but this is required for RAID and LVM install. Even if I try force option (--f), I still get above error. (Yes, this box has LVM) Any ideas? Cheers, Daniel.

Matthew Cengia <mattcen(a)gmail.com> writes: IMO it's a bad idea to add unnecessary complications to /boot regardless of what bootloader you're using -- because when it inevitably breaks, the jigsaw puzzle is easier.

On Sat, Oct 26, 2013 at 05:21:13PM +1100, Tony White wrote: it's not an ide issue or a raid issue, it's a driver issue. specifically, the aacraid apdatec raid controller driver. if you used this card as a hardware raid controller (rather than as JBOD ports) then you will need the aacraid driver loaded before linux will recognise it. this driver needs to be on the initramfs, and the initramfs needs to know that it should load the driver before trying to mount the root filesystem. you'll need to boot a rescue cd or similar - if you still have it, your original IDE drive would make an ideal rescue system as it has the exact same version of the kernel as on the adaptec raid drives. what kernel version are you running, and what distro? re-reading the Subject line, it appears you're using centos 5. kernel version? anyway, you'll need to boot your rescue cd / disk, load the aacraid module, find and mount the root (and /boot if you have one) filesystems, bind-mount /proc, /dev, and /sys, chroot into them, use a distro-specific method to force the initramfs to load the aacraid driver (e.g. on debian, the simplest way is to add 'aacraid' to /etc/modules and run 'update-initramfs -u -k all'), exit the chroot, unmount the filesystems, and reboot. for many reasons (including avoidance of proprietary lock-in - you're now stuck with adaptec controllers) you probably would have been better off using linux's built-in software raid, mdadm - the only time hardware raid is better is when you're using raid-5 or raid-6 and the card provides non-volatile write cache (so that raid5/6 write performance doesn't suck). and even then you're better off with zfs than raid-5. since the card only supports raid1 and raid10, that's not a factor. according to http://www.adaptec.com/en-us/products/series/2/ the card has some sort of hybrid ssd+hdd raid mode, but i can't see any details on what that actually means - in any case, zfs does it better, and without proprietary lock-in (i.e. with any drives on any controllers). amongst other benefits, converting from a single old IDE drive to mdadm raid would have "just worked" because it wouldn't have required any special drivers. huh? i thought you said it was IDE? for mdadm software raid, you could have just plugged these into the motherboard's SATA ports and avoided the expense and hassle of the adaptec card. if all else fails, you can always fall back to that. third-party tools like this aren't needed on linux - rsync will do the job, just as well, and for free. this almost certainly confirms that the aacraid driver was not loaded, so the linux kernel can't see the raid array containing the root filesystem and therefore can't find the rootfs. craig -- craig sanders <cas(a)taz.net.au>