On 8 Jun 2000, Chmouel Boudjnah wrote:
> > The machine is a Dual P6 600
> > 512M Ram
> > Mylex DAC960 Raid controller
> > 3 - 18Gb Drives
> > All partitions using ReiserFS (except swap)
> > Boot loader LILO
> > This was a new Install via Expert Mode (server)
> >
> > The message above comes up after loading the Mylex Module
> > Any Ideas how to fix it? Since this was a fresh install I'm not sure
> > re-installing is going to do any good?
>
> don't use ReiserFS with Mylex 8\.
I also reported this bug before the beta3 tree was frozen (or so I think).
(and yes reiserfs+mylex works fine - just not when the mylex driver is
loaded as a module - I have 2 servers running it at this time witn no
problems ;)
Here's what I did:
1. Partitioned my disks as follows (see EOL for mount points -- but you
don't want to enter the current mount points into diskdrake):
Disk /dev/rd/c0d0: 255 heads, 63 sectors, 6699 cylinders
Units = cylinders of 16065 * 512 bytes
Device Boot Start End Blocks Id System
/dev/rd/c0d0p1 * 1 65 522081 83 Linux /
/dev/rd/c0d0p2 66 195 1044225 83 Linux /var
/dev/rd/c0d0p3 196 325 1044225 83 Linux /tmp
/dev/rd/c0d0p4 326 6699 51199155 85 Linux extended
/dev/rd/c0d0p5 326 1369 8385898+ 83 Linux /usr
/dev/rd/c0d0p6 1370 6438 40716711 83 Linux /db01
/dev/rd/c0d0p7 6439 6699 2096451 82 Linux swap
(This is an oracle database server....i would have prefered to use
something else like mysql or postgresql but these people refuse to
learn anything new =/)
/db01 would normally be /home if this was, say, a shell server or
something -- do what you need to with the partitions, this is my
example ;)
2) took the /db01 space and created a 4g partition at te beginning, tagged
the mount point /, type is ext2.
3) install your system as if you were installing it normally.
4) when the system comes up (it will this time), rebuild the kernel with
your DAC960 and ReiserFS options set to "Y" (NOT "M"). (If you're
using menuconfig to configure, it's "*"). Set your other options, make
sure you turn SMP on with APIC so you get both processors when you boot
the new kernel =)
5) add the new kernel to /etc/lilo.conf and re-run lilo (I did not use a
ramdisk, here's my /etc/lilo.conf entry):
image=/boot/vmlinuz-2.2.15-bci
label=linux
root=/dev/rd/c0d0p1
append=" mem=1024M"
read-only
6) reboot your system with the new kernel (You relabeled the old one as a
backup didn't you?)
7) mkreiserfs on the unused partitions I tagged for /, /var, /tmp, and
/usr, then mount them all under /mnt/disk in the proper heirarchy
(remember to use -notail for speed, if you choose not to use -notail
you at least must use it on the partition /boot resides on or lilo
won't work.)
8) cd /; ls -a , then
cp -a (everything except /mnt /proc /lost+found) /mnt/disk
note if you do cp -a . /mnt/disk or cp -a / /mnt/disk you'll end up
copying /mnt/disk/ over and over and over and over and over.. ... ...
9) fix /mnt/disk/etc/fstab to reflect your future mountpoints.
10) fix /mnt/etc/lilo.conf to reflect your new / filesystem (make sure not
to change the root on your backup kernel image or you won't stand a
chance if something was wrong in the kernel config you just built)
11) run 'lilo -r /mnt/disk' if it says something about holes you forgot
to use -notail as a mount option on your /boot filesystem.
12) reboot. All should be well....if it's not, don't blame me because I
only tested on two machines with identical hardware from Penguin
Computing =)
13) you proabably want to look thru /var/log/boot.log and the output
from "df" and "mount" to make sure the reiserfs partitions are
mounted -- the old ext2 partition shouldn't be there anywhere unless
you did something in the new fstab to put it somehere.
14) once you're up and running the your real filesystems, fdisk and delete
the old ext2 / you installed on. You can then make that partition
the full size and mkreiserfs on it (you probably need to reboot before
running mkreiserfs because linux still can't re-read the partition
table on the fly when it has mounted partitions.)
15) Find a "certified" solaris admin to watch while you hit the reset
button without shutting down cleanly and see their amazement when your
machine that costs 1/4 as much as a solaris box with 1/2 the power
boots with dirty filesystems in 1/2 the time their E10K takes ;)
-dwild
