[BBLISA] Help with Dell PE2950 and RAID5 setup

Eddy Harvey bblisa2 at nedharvey.com
Thu Mar 8 08:39:56 EST 2007


It just so happens I use CentOS 4.4 x86_64, on poweredge 2950's, with perc5
controllers and 6-disk raid5.  I've spent a lot of time working on all this
stuff, and specifically on grub.  :-)  So you've got a bonafide expert this
time.  FWIW, I feel it is best practice to build the raid5 using 5 disks,
and leave the 6th as hotspare.  Also FWIW, I feel it is best practice to
make the / partition not an LVM, and make the swap partition not an LVM, but
the rest of the disk should be one PhysicalVolume, and one VolumeGroup.  All
other partitions should be their own LogicalVolume because it costs no
overhead, but gives you more flexibility someday down the road when you
discover you might want to change the size of it.  In particular, /home and
/var have a tendency to grow.

I can't imagine any reason you would experience what you experienced.  When
you install the OS, be sure to verify all 4 CD's.  Make sure the test says
"PASS" for each one.  I've never experienced what you're talking about on
any of my PE 2950's.

That being said, the way I would approach fixing it is...  
I am guessing the installer incorrectly installed grub for you, and
incorrectly configured the grub.conf.

You have already fixed grub, by doing your root and setup commands; this is
evidenced because you are now able to launch the kernel.

Now you have to fix grub.conf.

Insert CD1.  At the boot prompt type "linux rescue"  Yes, mount the proper
distribution, etc etc.

At the command prompt, chroot /mnt/sysimage (or whatever the instructions
say to do.)

Run grub.
Grub>  find /grub/stage1     (this is your /boot partition)
For the sake of this email, I assume it says:    hd0,3

Grub>  find /etc/hosts       (this is your / partition)
For the sake of this email, I assume it says:    hd0,1

Grub>  root (hd0,3)        (just to be sure)
Grub>  setup (hd0)         (just to be sure)
Quit.

Now, make sure to edit /etc/grub.conf, and be sure it also has correct
root=hd0,1 in there.  Inside your grub.conf file, when you say root=, you
are not specifiying the /boot partition, you're specifying the actual /
partition.

In the future, there is no reason to ever separate the /boot partition.
That was required for 486's, whose bios could not boot from a partition if
the partition extended past cyllinder 1024.  This is archaic, and only
causes extra confusion nowadays.



-----Original Message-----
From: bblisa-bounces at bblisa.org [mailto:bblisa-bounces at bblisa.org] On Behalf
Of Scott R Ehrlich
Sent: Wednesday, March 07, 2007 9:51 PM
To: bblisa at bblisa.org
Subject: [BBLISA] Help with Dell PE2950 and RAID5 setup

I've got a new PE2950 rack system with PERC controller and 6 drives set up
contiguously as one RAID5 volume on the PERC controller.

Using 64-bit CentOS 4.4, I partitioned for /boot having 100 Meg, swap having
2 gig, and / with the rest (removed LVM).

After the install was complete, the system rebooted and just gave me a
broken grub prompt (not the graphical Grub boot selector, but grub>).

I fed it root (hd0,0) and setup (hd0) and rebooted.

I then got the grub gui, but also got the following after the attempted
kernel boot:

Decompressing Linux...done.
Booting the kernel.
Red Hat nash version 4.2.1.8 starting
sda: asking for cache data failed
sda: assuming drive cache: write through
sda: asking for cache data failed
sda: assumint drive cache: write through EXT3-fs error (device sda3):
ext3_get_inode_loc: unable to read inode block - inode=2132343,
block=1231234
mount: error 5 mounting none
EXT3-fs error (device sda3): ext3_get_inode_loc: unable to read inode block
- inode=123123, block=124324123 EXT3-fs error (device sda3):
ext3_get_inode_loc: unable to read inode block - inode=6353565,
block=34634635
WARNING: can't access (null)
exec of init ((null)) failed!!!: 14
EXT3-fs error (device sda3): ext3_get_inode_loc: unable to read inode block
- inode=1245546546, block=645243235

umount /initrd/dev failed: 5
Kernel panic - not syncing: Attempted to kill init!


What might be wrong?   Bad hard drive?   This, again, is a fresh install,
straight off the 64-bit CentOS 4.4 4 CD set.

Thanks.

Scott

_______________________________________________
bblisa mailing list
bblisa at bblisa.org
http://www.bblisa.org/mailman/listinfo/bblisa




More information about the bblisa mailing list