Re: Boarding SuSE with Debian

2002-08-14 Thread Adrian 'Dagurashibanipal' von Bidder
On Thu, 2002-08-08 at 15:13, Michael Holzt wrote:
[installing Debian from the outside]

Yo!

I'll be doing exactly this in a while, I have already tested this
procedure at home and am just waiting until I can do it for real. Your
guidelines seem sensible, however, I feel I did it even easier:

1. create a tar from a fully working Debian system, as described in your
posting.
2. Instead of unpacking in a directory, I disabled swap on the target
system and unpacked into the swap partition. 
3. Recheck you got everything right: kernel, fstab, networking,...
4. install lilo to load the target system from the swap partition
5. boot
6. clean the SuSE away, copy everything from the 'swap' partition to the
target partition.
7. change partitions in fstab and lilo.conf, reinstall lilo
8. boot
9. reenable swap, start installing additional packages and services etc.

I feel this procedure is 'cleaner' because I really can mkfs in the
procedure (changing from ext2 to reiser in the process). If I'm careful
I can even fdisk and change pretty much everything I like (Additional
reboot after fdisk!).

The Debian base system was a 25M tar.bz2 file, sufficiently small to
upload through my 64k line. 

In any case: test on a similar system at home first! Make sure you know
wat you are doing!

(Hmmm... somebody should write this scripted and spread it as a Code-Red
like worm Let the Debian worm install Debian on all infected IIS
installations... :-] )

cheers
-- vbi

-- 
secure email with gpg http://fortytwo.ch/gpg


signature.asc
Description: This is a digitally signed message part


Re: Boarding SuSE with Debian

2002-08-14 Thread Markus Werner
Hello,

a little correction:-)

 7. In SuSE's lilo.conf add a section with Debian's boot and set it
as default (but still leave SuSE section).
Leave suse as default. Run lilo and afterwards lilo -R debianimage.
Reboot.

Go throu Step 8 till 13.

 14. If it doesn't come up -- ask the ISP to go to the console, reset
 the machine and choose SuSE at lilo prompt. I hope they can
 do it for free.
In Step 14 just ask for a Reset.


cu

markus






Re: Boarding SuSE with Debian

2002-08-08 Thread Nate Campi
On Thu, Aug 08, 2002 at 12:57:24PM +1000, Donovan Baarda wrote:
 On Wed, Aug 07, 2002 at 03:24:59PM -0700, Nate Campi wrote:
  On Wed, Aug 07, 2002 at 03:17:45PM -0700, Nate Campi wrote:
   
   I'd recommend using the swap
   partition as suggested above, and using the original kernel from suse.
   You know it boots the box, so use it again.
  
  Er, I meant the _currently_ running kernel. Using a known good kernel
  cuts down the number of things that can go wrong - the guiding principle
  for sysadmins ;)
 
 Yes, but be _very_ careful... you will need more than just the kernel, you
 will also need any initrd images, and you will need to make sure that the
 kernel+initrd will work for the new configuration... ie don't change fs type
 or anything.
 
 In some ways trying to use the same kernel introduces as much risk as it
 avoids... I'd probably be inclined to risk a new kernel rather than risk
 stuffing up running the old kernel under a new distro...

There's always a hundred reasons one approach *could* fail, and if you
have an old kernel that needs modules and an initrd image this approach
is more difficult.

He needs to be _very_ careful with this approach no matter which advice
he acts on. If his current kernel is a non-modular kernel with no initrd
image (like mine was), I strongly advise re-using it.
-- 
Yea, tho I walk thru the valley of the shadow of clues, I shall fear
no luser, for Thou lart with me, Thy chicken and Thy manual, they
comfort me. 



pgpS7sxoIiaTf.pgp
Description: PGP signature


Re: Boarding SuSE with Debian

2002-08-08 Thread Donovan Baarda
On Wed, Aug 07, 2002 at 09:19:28PM -0700, Nate Campi wrote:
 On Thu, Aug 08, 2002 at 12:57:24PM +1000, Donovan Baarda wrote:
  On Wed, Aug 07, 2002 at 03:24:59PM -0700, Nate Campi wrote:
   On Wed, Aug 07, 2002 at 03:17:45PM -0700, Nate Campi wrote:
[...]
  In some ways trying to use the same kernel introduces as much risk as it
  avoids... I'd probably be inclined to risk a new kernel rather than risk
  stuffing up running the old kernel under a new distro...
 
 There's always a hundred reasons one approach *could* fail, and if you
 have an old kernel that needs modules and an initrd image this approach
 is more difficult.
 
 He needs to be _very_ careful with this approach no matter which advice
 he acts on. If his current kernel is a non-modular kernel with no initrd
 image (like mine was), I strongly advise re-using it.

Agreed. _If_ the existing kernel is non-modular with no initrd, then use it.
You still want to make sure you are using a fs that is compiled in.

However, most distro's now use fully modular initrd images. I have found the
Debian default modular kernels to work fine with every hardware combination
I have thrown at them. 

The only time I compile kernels now is when I want something highly unusual
and/or special that the default kernels don't have (ie, DOV, IPSec, etc).
Even in these cases, I use the same kernel config, only tweaking it as
required, and build a kernel-package modular initrd packages. This way I
know that the kernel can handle any fs or additional hardware I throw at it
without having to recompile.


-- 
--
ABO: finger [EMAIL PROTECTED] for more info, including pgp key
--




Re: Boarding SuSE with Debian

2002-08-08 Thread Henning Heyne
If you have only one big partition for suse, maybe you can use parted (you can 
search for it at freshmeat) or other tools to reduce the partition. Then you 
can follow the steps below.. but try the whole process on a local machine.

On Wednesday 07 August 2002 18:15, Marcin Sochacki wrote:
 On Wed, Aug 07, 2002 at 07:55:26PM +0200, Florian Bantner wrote:
  my mission is following: Have rented a cheap server from
  an cheap hoster for a customer of ours. Only drawback: It
  is running suse linux. Since the provider is so cheap, he
  tells us: Do with the server what you want. And so I want
  Debian to take over. The problem:
 
  1. No access to neither floppy nor cd-rom
  2. Only service I get is pressing the reset button
  3. The other service is to reinstall base suse-
 installation if the server fails to come up.
 This will cost me 70 EUR/USD
 
  What is the best way to get Debian on this box? And how
  can I avoid the reboot-fail-reinstall-pay-tray again
  trap?

 If the machine has a secondary HDD or unused partition, or at least
 two partitions, you can install Debian parallel to the existing SuSE.

 1. Install a minimal Debian system on some local box. Remember to set
everything (kernel modules, IP address, etc.) just as you would do
on the server.
 2. tar.gz the whole installation into one big archive (ommiting /proc).
 3. Copy the archive to the server.
 4. Create a new filesystem on spare partition/disk (or if SuSE already
occupies several partitions move the data, so that one of the
partitions is free).
 5. Untar the archive to the fresh filesystem.
 6. Correct the entries in Debian's /etc/fstab to match those on remote
server.
 7. In SuSE's lilo.conf add a section with Debian's boot and set it
as default (but still leave SuSE section).
 8. Make sure you have prompt and timeout options in lilo.conf.
 9. Run lilo.
 10. Examine the Debian setup carefully, again.
 11. Reboot the machine.
 12. Pray/drink a coffee while pinging the machine.
 13. If it comes up -- you have a Debian system and gradually you can
 remove SuSE.
 14. If it doesn't come up -- ask the ISP to go to the console, reset
 the machine and choose SuSE at lilo prompt. I hope they can
 do it for free.

 Some caveats: /proc filesystem, swap partition, lilo, fstab, sshd/telnetd.

 Good luck!
 Marcin




Re: Boarding SuSE with Debian

2002-08-08 Thread Michael Holzt
I'm not subscribed to this list, so i hope my message gets through.
I heard about this Thread in IRC.

I managed in the past to convert a 11 Root-Server (a cheap server
product here in germany) on the fly from SuSE to Debian. In theory
you could just boot the Rescue-System (which will start some kind
of a Debian over the Network into a ram disk) and replace the
contest - you cannot destroy the rescue system so no problem here.

But when i attempted to convert, the rescue system was down for
maintenance and i didn't wanted to wait. So i've gone the hard
approach which in fact wasn't that hard anyway.

So here is what i successfully did: I installed a debian on a spare
hard disk at home. I compiled a kernel for the needs as on the 
target system and i build a new kernel with network card support
built in (not module - just to be sure). I configured everything so
the system should come up cleanly.

--- short interruption for people with 11 Root-Servers: ---

  One point here for people who might read this and want to try this
  with a 11 Root-Server: 11 has a strange network configuration
  with a netmask of /32 (=255.255.255.255). The preinstalled SuSE is
  configured with DHCP, but i strongly recommend going a approach with
  static configured IP. At least pump will NOT accept this config and
  your machine will fail to come up. But the static approach is not
  without tricks too, you need to use the pointopoint keyword i a
  way like this:

  iface eth0 inet static
address 217.160.107.39
netmask 255.255.255.255
gateway 217.160.107.1
broadcast 217.160.107.39
pointopoint 217.160.107.1

--- now back to the generic approach ---

You should configure the debian to a point where everything works
as expected. Don't forget to install SSHd, and add a user from
which you can login! Otherwise: FUBAR.

After i setup the system, and rebooted into it to make sure that
the network configuration will come up as it should in the target
system, i made a .tar.bz2 from the System (don't add /proc, /mnt,
/floppy, /cdrom - create them on the target system - but don't
forget that!!!).

I transfered the .tar.bz onto the target system and unpacked it
into a subdirectory of /, e.g. /debian. In the meanwhile i compiled
a static linked ls and mv, and copied them to the system into a
directory /stbin. Having statically linked binaries allows you to
move the libs out of the way without beeing fucked after that.

Finally i made:

  cd /
  mkdir /suse
  /stbin/mv home bin dev etc lib opt quota.user root sbin user var /suse

This will move everything except /mnt, /proc etc. out of the way

  /stbin/mv /debian/* /

This will move my Debian System onto the normal locations. Now you
should be able again to use normal ls, mv and other commands. You
will now need to correct fstab and mtab, just take them from SuSE:

 cp /suse/etc/fstab /suse/etc/mtab /etc


Now check if everything looks ok. Run LILO - however if you forget
this, it should boot the SuSE kernel, which is still on the same
position on the harddisk, just in a different directory entry, so
this should not be a real desaster.

Keep your Fingers crossed - and reboot the system.

Worked for me, but nothing for the faint of heart. If you have
a rescue system like at 11: Use it.
 
-- 
Greetings
Michael




Re: Boarding SuSE with Debian

2002-08-07 Thread Marcin Sochacki
On Wed, Aug 07, 2002 at 07:55:26PM +0200, Florian Bantner wrote:
 my mission is following: Have rented a cheap server from
 an cheap hoster for a customer of ours. Only drawback: It
 is running suse linux. Since the provider is so cheap, he
 tells us: Do with the server what you want. And so I want
 Debian to take over. The problem:
 
   1. No access to neither floppy nor cd-rom
   2. Only service I get is pressing the reset button
   3. The other service is to reinstall base suse-
  installation if the server fails to come up. 
  This will cost me 70 EUR/USD
 
 What is the best way to get Debian on this box? And how
 can I avoid the reboot-fail-reinstall-pay-tray again
 trap?

If the machine has a secondary HDD or unused partition, or at least
two partitions, you can install Debian parallel to the existing SuSE.

1. Install a minimal Debian system on some local box. Remember to set
   everything (kernel modules, IP address, etc.) just as you would do
   on the server.
2. tar.gz the whole installation into one big archive (ommiting /proc).
3. Copy the archive to the server.
4. Create a new filesystem on spare partition/disk (or if SuSE already
   occupies several partitions move the data, so that one of the
   partitions is free).
5. Untar the archive to the fresh filesystem.
6. Correct the entries in Debian's /etc/fstab to match those on remote
   server.
7. In SuSE's lilo.conf add a section with Debian's boot and set it
   as default (but still leave SuSE section).
8. Make sure you have prompt and timeout options in lilo.conf.
9. Run lilo.
10. Examine the Debian setup carefully, again.
11. Reboot the machine.
12. Pray/drink a coffee while pinging the machine.
13. If it comes up -- you have a Debian system and gradually you can
remove SuSE.
14. If it doesn't come up -- ask the ISP to go to the console, reset
the machine and choose SuSE at lilo prompt. I hope they can
do it for free.

Some caveats: /proc filesystem, swap partition, lilo, fstab, sshd/telnetd.

Good luck!
Marcin




Re: Boarding SuSE with Debian

2002-08-07 Thread Joey Hess
Florian Bantner wrote:
 Ahoy friendly Debian fellows,
 
 my mission is following: Have rented a cheap server from
 an cheap hoster for a customer of ours. Only drawback: It
 is running suse linux. Since the provider is so cheap, he
 tells us: Do with the server what you want. And so I want
 Debian to take over. The problem:
 
   1. No access to neither floppy nor cd-rom
   2. Only service I get is pressing the reset button
   3. The other service is to reinstall base suse-
  installation if the server fails to come up. 
  This will cost me 70 EUR/USD
 
 What is the best way to get Debian on this box? And how
 can I avoid the reboot-fail-reinstall-pay-tray again
 trap?

Install debian in a chroot with debootstrap. If you have a spare
partition to use as the chroot you can eventually make it the bootalbe
partition.

-- 
see shy jo




Re: Boarding SuSE with Debian

2002-08-07 Thread Emilio Brambilla
hello,
On Wed, 7 Aug 2002, Marcin Sochacki wrote:

[...]
 7. In SuSE's lilo.conf add a section with Debian's boot and set it
as default (but still leave SuSE section).
 8. Make sure you have prompt and timeout options in lilo.conf.
 9. Run lilo.
using lilo -R may help you at point 14

 14. If it doesn't come up -- ask the ISP to go to the console, reset
 the machine and choose SuSE at lilo prompt. I hope they can
 do it for free.
you will ask him only to reboot the server, and it will boot suse
automagically (as it will be the second boot)

-- 
bye,
emilio




Re: Boarding SuSE with Debian

2002-08-07 Thread Marcin Sochacki
On Wed, Aug 07, 2002 at 08:46:41PM +0200, Emilio Brambilla wrote:
 using lilo -R may help you at point 14

That's a brilliant suggestion. Thanks!

Marcin




Re: Boarding SuSE with Debian

2002-08-07 Thread Florian Bantner
On Mit, 07 Aug 2002, Marcin Sochacki wrote:

 On Wed, Aug 07, 2002 at 07:55:26PM +0200, Florian Bantner wrote:
  my mission is following: Have rented a cheap server from
  an cheap hoster for a customer of ours. Only drawback: It
  is running suse linux. Since the provider is so cheap, he
  tells us: Do with the server what you want. And so I want
  Debian to take over. The problem:
  
  1. No access to neither floppy nor cd-rom
  2. Only service I get is pressing the reset button
  3. The other service is to reinstall base suse-
 installation if the server fails to come up. 
 This will cost me 70 EUR/USD
  
  What is the best way to get Debian on this box? And how
  can I avoid the reboot-fail-reinstall-pay-tray again
  trap?
 
 If the machine has a secondary HDD or unused partition, or at least
 two partitions, you can install Debian parallel to the existing SuSE.
 
 1. Install a minimal Debian system on some local box. Remember to set
everything (kernel modules, IP address, etc.) just as you would do
on the server.
 2. tar.gz the whole installation into one big archive (ommiting /proc).
 3. Copy the archive to the server.
 4. Create a new filesystem on spare partition/disk (or if SuSE already
occupies several partitions move the data, so that one of the
partitions is free).
 5. Untar the archive to the fresh filesystem.
 6. Correct the entries in Debian's /etc/fstab to match those on remote
server.
 7. In SuSE's lilo.conf add a section with Debian's boot and set it
as default (but still leave SuSE section).
 8. Make sure you have prompt and timeout options in lilo.conf.
 9. Run lilo.
 10. Examine the Debian setup carefully, again.
 11. Reboot the machine.
 12. Pray/drink a coffee while pinging the machine.
 13. If it comes up -- you have a Debian system and gradually you can
 remove SuSE.
 14. If it doesn't come up -- ask the ISP to go to the console, reset
 the machine and choose SuSE at lilo prompt. I hope they can
 do it for free.
 
 Some caveats: /proc filesystem, swap partition, lilo, fstab, sshd/telnetd.
 
 Good luck!
 Marcin
 
 
 -- 
 To UNSUBSCRIBE, email to [EMAIL PROTECTED]
 with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
 

Thanks for the answer (to the other, too), but still there are two
points:
1. Only one big partition (30GB) mountet as /
2. Noone will go to the terminal for me, which means
   either it comes up or -- hello again plain SuSE, goodbye 70.-

Best regards

Florian

-- 
--
Florian Bantner  AXON-E interaktive medien
Tel. +49-941-599 854 4  Fax. +49-941-599 854 1
Mail [EMAIL PROTECTED]
Key  http://www.axon-e.de/gpg/f.bantner.key
1191 0C87 D9DB 3217 ABBA  5223 6D74 AB19 5C9D FC49
--




Re: Boarding SuSE with Debian

2002-08-07 Thread Nate Campi
On Thu, Aug 08, 2002 at 12:00:25AM +0200, Marcin Sochacki wrote:
 On Wed, Aug 07, 2002 at 09:21:45PM +0200, Florian Bantner wrote:
  Thanks for the answer (to the other, too), but still there are two
  points:
  1. Only one big partition (30GB) mountet as /
  2. Noone will go to the terminal for me, which means
 either it comes up or -- hello again plain SuSE, goodbye 70.-
 
 But wait, maybe there is a separate swap partition configured?
 You could switch off the swap and use that partition for Debian.
 I guess 64 MB could be enough. In case you have very little RAM available,
 you could use a regular swap file.

Just two days ago I converted a host in a datacenter to debian from
redhat, from the comfort of my home. I'd recommend using the swap
partition as suggested above, and using the original kernel from suse.
You know it boots the box, so use it again. Use lilo's chroot (-r)
option to install lilo from within the new debian filesystem after you
copy over a good install from another woody box. Use the -v and -t
options to see what it will do before running lilo for real on the new
install.

I cloned a woody box like this:

# rsync -avz -e ssh --exclude 'proc' / susebox:/swapmount

Don't copy /proc - weird things happen, make sure /proc exists on the
new filesystem though (mkdir /proc after the rsync).

I'm leaving out some details, but I think they were all covered in
previous messages in this thread.
-- 
Guide to understanding a net.addict's day: Slow day: didn't have much
to do, so spent three hours on usenet. Busy day: managed to work in
three hours of usenet. Bad day: barely squeezed in three hours of
usenet. - Unknown



pgpCFSmVUFLlP.pgp
Description: PGP signature


Re: Boarding SuSE with Debian

2002-08-07 Thread Nate Campi
On Wed, Aug 07, 2002 at 03:17:45PM -0700, Nate Campi wrote:
 
 I'd recommend using the swap
 partition as suggested above, and using the original kernel from suse.
 You know it boots the box, so use it again.

Er, I meant the _currently_ running kernel. Using a known good kernel
cuts down the number of things that can go wrong - the guiding principle
for sysadmins ;)
-- 
Man is still the most extraordinary computer of all. - JOHN F.
KENNEDY, speech (1963)



pgpiQU5iOV8NJ.pgp
Description: PGP signature


Re: Boarding SuSE with Debian

2002-08-07 Thread Donovan Baarda
On Wed, Aug 07, 2002 at 03:24:59PM -0700, Nate Campi wrote:
 On Wed, Aug 07, 2002 at 03:17:45PM -0700, Nate Campi wrote:
  
  I'd recommend using the swap
  partition as suggested above, and using the original kernel from suse.
  You know it boots the box, so use it again.
 
 Er, I meant the _currently_ running kernel. Using a known good kernel
 cuts down the number of things that can go wrong - the guiding principle
 for sysadmins ;)

Yes, but be _very_ careful... you will need more than just the kernel, you
will also need any initrd images, and you will need to make sure that the
kernel+initrd will work for the new configuration... ie don't change fs type
or anything.

In some ways trying to use the same kernel introduces as much risk as it
avoids... I'd probably be inclined to risk a new kernel rather than risk
stuffing up running the old kernel under a new distro...

-- 
--
ABO: finger [EMAIL PROTECTED] for more info, including pgp key
--