[SOLVED] GRUB configuration problem

[jenonen]

I am trying to configure a system to boot Windows XP, CentOS 4 and RHEL5. I have one hard drive that contains both Windows XP and CentOS 4, and a separate drive that contains RHEL5. Until recently, I only had one SATA cable, so I could only connect one drive at a time. Under this configuration, everything works fine. When the RHEL5 drive is connected, I can boot into it. When the Windows/CentOS drive is connected, I can dual-boot into either OS. (GRUB was configured on this drive automatically when I installed CentOS into a new partition.)

Opening the box and moving the SATA cable is a lot of trouble, so I finally got a second SATA cable and enabled both SATA0 and SATA1 in the BIOS. I currently have the Windows/Centos drive as the primary, and I can still boot into both Windows/Centos. Now, I want to add RHEL5 to menu, but I can't find the file GRUB is using to present its menu at startup.

I have configured GRUB before on other systems, but I just know the very basics, such as where the grub.conf file should be. So, I spent a whole day reading advice online and asking friends who might have experience with these issues. Here are the steps I have taken so far:

I confirmed there is no /boot/grub directory, and /etc/grub.conf is a broken soft-link to /boot/grub/grub.conf. I did a find for grub.conf, which found nothing. I did a find for menu.lst, which found one item -- an example GRUB config file in /usr/share/doc/grub-0.95. I noticed that when CentOS boots, I see the GRUB commands printed to the screen, the first of which is:

root (hd0,2)

So, I did a grep -R "(hd0" * at the / directory, which also found only one item -- the example menu.lst file in /usr/share/doc/grub-0.95. I discovered that I can go to the command line grub from the grub menu and do:

cat (hd0,2)/grub/grub.conf

The cat command returns a printout of the grub configuration the system is obviously using. I didn't create this file, but the titles are identical to what I see in the GRUB menu, the default boot is Windows, and the timeout is very short. This must be the file. It looks like:

default=2
timeout=5
splashimage=(hd0,2)/grub/splash.xpm.gz
hiddenmenu
title CentOS (2.6.9-89.ELsmp)
root (hd0,2)
kernel /vmlinuz-2.6.9-89.ELsmp ro root=/dev/VolGroup00/LogVol00 rhgb quiet
initrd /initrd-2.6.9-89.ELsmp.img
title CentOS-up (2.6.9-89.EL)
root (hd0,2)
kernel /vmlinuz-2.6.9-89.EL ro root=/dev/VolGroup00/LogVol00 rhgb quiet
initrd /initrd-2.6.9-89.EL.img
title Windows XP 32-bit
rootnoverify (hd0,1)
chainloader+1


However, when I am booted into CentOS, which is supposedly (hd0,2), there are no /boot/grub/grub.conf or /grub/grub.conf files to edit.

I talked with a friend who has a successful triple-boot configuration who sent me the contents of her /boot/grub directory. I tried creating the /boot/grub directory and populating it with my friend's files. Then, I modified her grub.conf to be the current grub configuration (above) with modified title names so I can tell from the GRUB menu if it is using the old or the new file. After reboot, it is still using the old title names.

Although I don't think my GRUB is broken -- it works just fine for the two OS's it is configured for -- I tried to re-install GRUB by giving the command:

grub-install /dev/sda

That command gives an error that says:

/dev/mapper/VolGroup00-LogVol00 does not have any corresponding BIOS drive.

I'm not really sure what that means, and the Logical Volume Management utility doesn't give me any information that makes sense to me. I have seen other forum experts ask for the output of fdisk -l for problems related to grub and partitioning, so if it helps, my output is:

Disk /dev/sda: 80.0 GB, 80000000000 bytes
255 heads, 63 sectors/track, 9726 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

Device Boot Start End Blocks Id System
/dev/sda1 * 1 5 40131 de Dell Utility
/dev/sda2 6 5105 40965750 7 HPFS/NTFS
/dev/sda3 5106 5118 104422+ 83 Linux
/dev/sda4 5119 9726 37013760 5 Extended
/dev/sda5 5119 9726 37013728+ 8e Linux LVM

Disk /dev/sdb: 250.0 GB, 250059350016 bytes
255 heads, 63 sectors/track, 30401 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

Device Boot Start End Blocks Id System
/dev/sdb1 * 1 13 104391 83 Linux
/dev/sdb2 14 30401 244091610 8e Linux LVM


I've also tried making the RHEL5 drive the primary drive. In that case, I can modify the existing /boot/grub/grub.conf file and see my changes at the GRUB boot menu. However, I can't get Windows to boot in this configuration. I've done a lot of google searching on the topic and added map commands to make Windows think it is on the primary drive. But, I'm still unsuccessful on this front as well. I think I'm closer to solving the problem with Windows/CentOS as the primary. However, if you think I will have more success with RHEL5 as the primary drive, I can provide more details as to my current grub.conf on that drive in a later post.

Any advice you have is greatly appreciated. Thank you in advance!

[jenonen]

I finally located the grub.conf file my CentOS is using. Based on a forum post, I discovered it was stored on a different partition. After booting into CentOS, I mounted /dev/sda3. In this partition was a grub/grub.conf, which I successfully modified and saw the changes at reboot.

Now, I have one more problem. I took the GRUB commands from my RHEL5 /boot/grub/grub.conf file, which was automatically generated by the RHEL5 installer and successfully boots RHEL5 when the RHEL5 drive is the primary drive. These commands are:

root (hd0,0)
kernel /vmlinuz-2.6.18-53.el5 ro root=/dev/VolGroup00/LogVol00 rhgb quiet
initrd /initrd-2.6.18-53.el5.img

I changed the root command to refer to hd1 and placed these commands into a new entry in my CentOS grub.conf. The new entry looks like:

title RHEL5
root (hd1,0)
kernel /vmlinuz-2.6.18-53.el5 ro root=/dev/VolGroup00/LogVol00 rhgb quiet
initrd /initrd-2.6.18-53.el5.img

However, when I choose the RHEL5 from the menu, it acts like it is booting CentOS, but it eventually fails and lets me try to login to the shell.

Does anyone have any advice on what I should change the root= portion of the command to, or if there is something else I am doing wrong with this configuration?

Thanks again.

[devils casper]

Hi and Welcome !

You are using LVM in both OSes and as GRUB doesn't support LVM, installer has created separate /boot partitions for both distros. All configuration files of grub, kernels and initrd images are located in /boot partitions only.

Boot up CentOS and mount /dev/sdb1 partition. Its /boot partition of RHEL. Post the contents of its boot/grub/grub.conf file here.


Post the contents of grub.conf file of CentOS too. Its /dev/sda3.

[Lostfarmer]

Have never used LVM but something looks wrong:

title CentOS-up (2.6.9-89.EL)
root (hd0,2)
kernel /vmlinuz-2.6.9-89.EL ro root=/dev/VolGroup00/LogVol00 rhgb quiet

title RHEL5
root (hd1,0)
kernel /vmlinuz-2.6.18-53.el5 ro root=/dev/VolGroup00/LogVol00 rhgb quiet

both 'root=' are the same, and that might be a problem (?).

Another problem is when the hdd is moved to a different IDE channel the /etc/fstab is likely wrong.

Do let 'devils casper' give his option on this.

[helenk579]

Thanks you for the post.
Hi guys, Im a newbie. Nice to join this forum.

[jenonen]

LostFarmer, you are exactly right. Since I had originally set up the two drives separately, both CentOS and RHEL5 had installed on as VolGroup00/LogVol00. When I put the drives in the same system, there was a conflict. The output of lvscan showed that two volumes' UUIDs were conflicting and that they system was randomly choosing one of them (in my case CentOS) to load.

To fix it, I renamed the volume group RHEL5 was installed on using:

vgrename <UUID shown as conflicting from lvscan> VolGroup00_rhel5

I updated my grub.conf entry for RHEL5 to use VolGroup00_rhel5 instead of VolGroup00, and I added entries to /etc/fstab for VolGroup00_rhel5.

Then, I had to update links within the kernel image to point to VolGroup00_rhel5. Sounds hard, but I followed the directions I found on another site "Working with Multiple LVM Hard Disks under Fedora Core Linux", and I can now boot all three systems! Here are the steps I took to update the image:

mkdir /mnt/sdb1
mount /dev/sdb1 /mnt/sdb1 #(the boot partition of my rhel5 drive)
cd /mnt/sdb1
mkdir newinit
cd newinit
gunzip -c ../initrd-2.6.18-53.el5.img | cpio -idmv
# or whatever initrd version file you have

Then, I edited the "init" file to replace all instances of VolGroup00 with VolGroup00_rhel5. Repackage a new image. (First rename the original one out of the way to save it so you can restore it later if need be.) Then, still within the "newinit" directory, do:

find . | cpio --quiet -c -o > ../newinitrd
cd .. # we are now in /mnt/sdb1
gzip -9 < newinitrd > initrd-2.6.18-53.el5.img
# or whatever it was called before...EXACTLY
sync
cd /
unmount /mnt/sdb1
unmount /mnt/VolGroup00_rhel5/LogVol00
sync

Thanks everyone for your help!

[Lostfarmer]

Thanks for posting your solution, it may help others. I would have been lost, O I still am.