I have some comments that go beyond this one-time operation and 
address the issue of creating hardware redundancy. Those comments 
are based on a tip Seth gave me about transplanting an entire 
installation.

A) Change your machines to swappable HDs (for convenience and 
reducing down-time)

B) tar up the individual partitions or directories (whatever 
makes sense and is allowed by free disk space); see(*)
 If the system is running /proc and /var may not like to be 
tar'ed. I used the -T option to define the directories I wanted 
to save/avoid. (if you have a CD burner those *.tgz also make a 
great backup on CD of the entire system)

C) Go to a second machine and make a small (console-only) helper 
installation at the end of one of the swappable HDs; partition 
the rest (major) part of the HD to accept the original source 
installation.

D) Use that helper installation to un-tar the various .tgz into 
the appropriate partitions (that way you preserve the file 
system-hardware relationship w/o having to change fstab ...) 

E) Create a lilo_transplant.conf that translates directory names 
to the temporary mount points within that helper installation, 
e.g. /boot may be called /mnt/hda1 , /home maybe /mnt/hda7 , etc.
 Use 'lilo -C lilo_transplant.conf' within the helper 
installation to write the MBR.

F) Boot to the transplanted system and see if the new hardware 
needs minor configuration (my NICs did).

Now you have a functioning duplicate of the original server.
If you need redundancy for other servers just get one swappable 
HD for each, repeat those step above, and you can switch to any 
of the backup system in two minutes. 
 
(*) It may be wise separate more permanent data (the system 
files) from those that change all the time (e.g. /home) -- that 
way you can update your .tgz more easily and un-tar them onto
those spare HDs.

 - Horst (I am on 'digest' - so this maybe outdated).

 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

On Tue, 24 Jul 2001, Seth Cohn wrote:

> 30 minutes?
> 
> Don't even worry about bounces... the 4 hour rule
> is fine.
> 
> I'd make sure you have a good backup, in case of
> unforseen problems.  Something you can restore
> from baremetal if need be.
> 
> 
> 
> 
> --- Rob Hudson <[EMAIL PROTECTED]> wrote:
> > I have a server that I'd like to upgrade
> > (hardware & OS), and want the
> > downtime to be minimal.  I'm going to
> > unsubscribe from the various
> > mailing lists I'm on so they don't get bounced
> > with the message 'will
> > try again every 4 hours'.  I don't really have
> > a server to take over
> > in the interim.  I project I'll be down maybe
> > 30 minutes.
> > 
> > Any advice?
> > 
> > Thanks,
> > Rob
> > 
> 
> 
> __________________________________________________
> Do You Yahoo!?
> Make international calls for as low as $.04/minute with Yahoo! Messenger
> http://phonecard.yahoo.com/
> 


Reply via email to