[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: IBM eServer x series 206 RH-to-Debian migration



On Thu, Nov 09, 2006 at 02:34:46PM -0500, Lee Whalen wrote:
> so I am tasked with migrating all of our HeadRat servers over to
> Debian.
> 
>   My big question is, does Debian work with the Adaptec SATA HostRAID
>   controller that this IBM eServer x206 box has on it?  

Etch may.  Make an install media, boot, and see if what disks it sees.
Its faster than fining out ahead of time.

> 
> For the curious, my "grand scheme" and "backout plan" is to initially
> get the entire filesystem off the box via "tar cvf - -C / . | ssh
> user@othermachine gzip -c \> backupfile.tar.gz"
> 
> (obviously with the flags to preserve permissions, stay on the local
> filesystem, etc).  Once I have my 9 gig tarball located somewhere
> safe, format the drives, set up the array, and install a base Debian
> system.  Once I'm safely booted into the new Debian kernel, I'd create
> /mnt/oldsystem or whatever and untar the old system back to that
> directory.  I'd then apt-get install an app (say, Apache), then copy
> it's configuration files and volatile data back from /mnt/oldsystem.
> Rinse and repeat for the users, NFS mounts, fstab, ticket application,
> etc. and after a few hours I'll have my shiny new Debian system with
> all the old applications, data, and settings back on it.  If for some
> reason that doesn't work and I need to revert back to the old system,
> I'd repartition the drives as they were before, boot a LiveCD, mount
> the partitions correctly under /target of the boot CD, then untar the
> whole thing back across the network.  chroot to /target, run grub,
> reboot, and (ideally) I'd have the "old" system back in a pinch.
> 
>   For those of you who managed to read this far (go YOU!) does that
>   sound feasible, or am I doing far, far too much work?
 
Tread slowly and carefully.  Plan it out step by step.  See if the
install media sees the raid array (I guess sees one disk [the array],
many disks [the array disks without the raid part], or no disks [doesn't
like the controller at all]).  If it sees one disk, you're OK.  If it
sees many disks, you're still OK becase you can setup kernel software
raid during the install.  If it doesn't see any, you've got a problem.

This part you can do without much down time and without any alteration
to your setup.

For installing packages, do it step-by-step.  You need an editor long
before you need X or apache.

Once you get your new installation complete and your /mnt/oldsystem (I
personally use /var/local/bacup), watch the configurations.  Don't just
copy files over because they will likely be different.  Instead you will
need to copy select configuration options over.  They will be tediously
similar but different enough.

Read the installation manual and the debian-reference.  They will help
you get your mind around the differences between RH and Debian.

Your backup file will be much smaller if you use bzip2 instead of gzip.
Your 9 gig tarball may turn into small enough for a DVD.

I don't work in a production environment.  However, I wonder if you have
a cold-swap spare server onto which you can install debian and transfer
the data to work out the kinks, put that into production and then cycle
through round-robin style.  That would keep your production downtime
very minimal and minimize debian-learning-curve issues.

As far as too much work, its no more work than if one of your production
servers had died and you needed to install on a replacement.  Its just a
slightly different OS.  And more stable IMHO.

Go for it.

Doug.



Reply to: