Re: SCO 5.0.5 dd / clone

In article <20060110064709.64734@xxxxxxxxx>,
Tom Parsons <sconews@xxxxxxxxx> wrote:
>Otto enscribed:
>| Hi Bill,
>|
>| Thanks for the feedback.
>|
>| I fully agree with what you write HOWEVER, in my case the server is a
>| simple email server with almost all POP accounts and also a SAMBA file
>| server (again most of the data is static) so a "little" bit of loss is
>| acceptable. I have used this "Disaster Recovery" method successfully on
>| several of my customer sites (running Linux) and it is really a 30
>| second (no kidding!!) reboot to get the system going in the event of a
>| disk failure. BTW I have also implemented a Microlite BackupEdge backup
>| as well. If you use the Disaster Recovery of Microlite, it takes
>| approximately 2-4 hours to fully recover (hence the 30 second recovery
>| for a mail server) is required by my customer. I really did now want to
>| get into the merits or lack thereof of the method used - just a reason
>| or fix as to why the "dd" command works fine under same conditions with
>| LINUX and not woth SO 5.0.5. I will probably try Bela suggestion and
>| try with SCO 5.0.7 and see whether it is indeed a kernel problem or
>| not. Again much thanks for your time and suggestions. Rgds. Otto.

>Which begs the question (at least here), why are you going through all
>of these gymnastics when there are much better and more reliable solutions.

>Mirror the drive and install a hot spare. Let technology do the work
>instead of kludging 1980 style partial solutions.

I also gave Otto some solutions as he sent me email, and I assumed
that he had not posted to the 'net.

One thing I pointed out is that dd is going to copy things over and
over and over again - as so much never changes, where something
like rsync would work well.

The 'dd' approach seems to probably come from the Ghost approach
that many people use in the MS world. But MS and Linux solutions
don't always apply very well in real Unix systems.

I also suggested raid.

He never indicated his hardware or size of the system so who is
to know if the 3-4 hour Microlite recovery is based only on the
amount of data he has. If it is large then the 'dd' is also going
to take a lot of system resources - most of which will be wasted.

I also pointed out to Otto that at times Linux approaches can spell
disaster when applied to a pure Unix system. I had to physically
pull a Sun Netra [from our racks] to repair a system that the
client had Linux-ized - so it would be more familiar.

After a shutdown it would not reboot as it could not find it's home
directory or shell. And the system would not let you reinstall
the OS as there was one there.

I wound up pulling the unit, swapping drives, installing Solaris
on the new drive. Then it was a simple 'fsck' on the old drive
from the new drive,and then swap them back and set things up the
way they were intended.

As you and I well know there are so many minor [and sometimes
major] differences between Unix systems, and particularly Unix-like
systems, that you can shoot yourself in the hard-drive if you try
to use a universal approach.

Bill
--
Bill Vermillion - bv @ wjv . com
.

Relevant Pages