Hello cookers,
I'm a linux-admin by profession, so I can say I have a lot of
experience with different distros. Not with mandrake yet though,
so here is my story how I failed get my 8.0 installation right, and
had to go for RedHat 7.1 instead, much to my regret.
I have a system at home with two hardrives, with
a / on a raid1 and /usr, /opt, /var, /data2 on raid0 filesystems.
I had RedHat 6.1 running.
At first I thought the partioning/filesystem tool was nice and
logical, but it appeared to be extremely buggy where it comes to raid
systems.
It would let me make the /dev/md0 ... /dev/md4 systems, but then
at the formatting stage it gave an error about mount-points and
/dev/md0 not actually existing and bring me back to the
partitioning tool again.
I tried with the shell prompt under ctrl-alt-f2, and after many
tries I got it working:
First: it is an extremely limited shell: backspace does not work,
no completion, there is no editor, so when you need to edit a file
you must use sed, and something like "cat /etc/raidtab".
I made a /etc/raidtab with the right entries that way, I did
modprobe raid0 raid1 raid5, I made the right mountpoints in /mnt
(all in the ramdisk filesystem), ran raidstart -c /etc/raidtab /dev/md0
and after much trial and error finally got the install program so far
that it would believe my filesystems where there and it got to
installing packages in there.
Then the installer crashed completely, no numlock, nothing, right after
the last package got installed. No kernel or lilo was installed yet,
so I had to do it all over again including including the pain with the
raid stuff.
This time it did not crash, and after rebooting the computer came up
with a quite beautiful graphical lilo boot screen, and subsequently
with a nice graphical boot.
Then it went wrong again. The booting stopped with a kernel panic:
cannot mount root on device 09:00, which is the device number for
/dev/md0, my / on a raid1 device. Somehow the booting kernel
would not auto-run my raid-devices although it should. This of course
sucks majorly, because you absolutely cannot get into your system,
unless you get really creative.
So I got creative. I booted with the cdrom again, made the /etc/raidtab
by typing it in with 'cat /etc/raidtab' again, probed the modules, ran
the raid devices etc.etc.
Then I mounted the /boot partition (which is not raid of course), found
the right kernel there, copied it to a floppy, mounted the /usr filesystem
so I could get to the rdev program, and did a rdev /dev/fd0 /dev/md0.
Then I booted with the floppy, with exactly the same result: kernel panic.
Mandrake people: your kernel does not boot from raid when installed
with your installation software, and it should, otherwise the software
should have given me a warning about this.
And if the idea is that the kernel should load modules for raid first,
then it should have installed an initial ramdisk (initrd) to do this.
I finally fixed it by using the same floppy with
'rdev /dev/fd0 /dev/hda8' (/dev/hda8 being one half of my /dev/md0),
and of course not forgetting to change the /etc/fstab accordingly.
Now you try to do this all with the crippled shell you get. (it gave
all kinds of perl errors all the time, strange..)
Anyway, finally I got it to boot, and the filesystems /data2 and /home
that I wanted to spare were still there. Good.
Then I tried to get networking online, and I have a small internal
100Mbit network, and for internet I have an outgoing ISDN line.
It was a major pain to get this configured, and in the end I just gave
up and decided to install RedHat.
The drakconf tools all spoke for themselves, and I should have had a
working network.
However, nothing worked right:
The nfs-server would not work at all, I wanted to use this mandrake
computer so that my other Linux machine could mount its /home from this one.
I tried everything, disabled or enabled bastille-linux did not make a
difference.
SSH was the same. I could only ssh to the local loop, 127.0.0.1 , but
not to localhost. From my other computer I could not ssh to the mandrake,
could not even ping it.
The mandrake machine could ssh to the other however.
Something must have been wrong with the firewall but then it should have
worked when I shut that off.
Some more problems:
- almost all directories in / had the wrong permissions, so that non-root
users could not even run ls or anything else anymore.
For instance: /usr,/bin,/usr/bin,/var,/etc etc.
all where 'root adm rwx--x--x' .
I had to fix that by hand.
- Mandrake comes with ALSA, but there was no
clue how to actually get it working. If you ship it, there should be
some nice configuration program for it, or better yet, that should be
automatic.(especially for THE desktop linux)
I had to use OSS instead. I had no clue how to configure it, and when
configured, how to actually use it, and no time to find that all out.
I'm really disappointed that I have no time to try and fix all these
(and some more I didn't mention) problems with Mandrake 8.0. But I had
to get that machine up and running again, so I installed a RedHat 7.1
instead with none of these problems.
I'm confident that given enough time I could have solved all of the
problems above, but for a distribution that aims to be a ease-of-use
and easy-install, I think that the above shows that there were quite
some bugs in the installation that would have stopped the avarage
say Microsoft windows users from pursuing it any further.
Roland
--
