SUMMARY: "hardware level" Cloning IDE-disks for rapid deployment/backup/etc.

From: Tim Chipman <>
Date: Fri Mar 04 2005 - 09:42:46 EST
I will confess, this posting is both a question and an answer, all at 
once. I made this observation earlier today while testing an inexpensive 
PCI IDE Raid Controller card, intended for mirroring a windows box - and 
I gave this procedure a go "just for a lark", using a drive pulled from 
an Ultra5. The implications were big enough (and something I wish I had 
known .. a long time ago ..) that I thought it would be of interest to 
other subscribers of the list.

Q: How to make an exact duplicate of an IDE disk from a production 
system, in minimal time, which yields a perfect bootable drive identical 
to the original, suitable for backup/archive on shelf ; bulk deployment 
of many "identical/similar" systems ; etc.

My observed solution: Using an ultra5 with solaris8 installed on a 
WDC20gig HDD as the "source" ; a blank Maxtor 30gig drive as the 
"target"; and the steps outlined below, I ended up putting the Maxtor 
disk back in the ultra5 and booting up "the same system" - down to the 
level even that "format" reported I had a WDC 20gig part installed.

Steps were,

-Buy the appropriate IDE raid controller PCI Card (~$28 cdn. at my 
online supplier, see notes below), capable of raid0-1-10-jbod via 
single-chip controller..

-install to available PCI slot of a standard X86 system (pentium and up, 
probably). Note that there is no host OS involved in this procedure (we 
don't boot into dos/windows/linux/solaris)

-obtain a "target" IDE drive of equal or greater capacity to the 
"donor", jumper as master and attach to "secondary" bus on raid card.

-shutdown your sparc machine, remove IDE disk, attach to primary bus on 
Raid card. Ensure drive is jumpered as "master" if not already thus.

-boot the x86 board, enter raid-bios which appear after initial "POST", 
define mirror composed of 2 drives. When prompted, do not erase data in 
mirror ; instead indicate "source" and "target" drives for cloning. This 
proceeds at a rate of ~12-15 minutes for a 20 gig drive set ; 
approximate linear timeframe for larger/smaller drives. My drives were 
both 7200rpm parts & ATA100 capable, I suspect "lower performance" disks 
might be a bit slower at this step. I don't believe it makes any 
difference how "full" the donor disk was to begin with.

-once completed, mirror status is shown as "functional". Shutdown and 
remove IDE drives, put the original on the shelf and install the 
"target" back in the original host.

If you wanted to do this "more than occasionally", I suppose generic 
IDE-drive trays could be used to simplify/"neaten" the process of 
temporarily attaching the source and target ide disks onto the raid 
card. In my case, I did initial testing with a bare Celeron800 
motherboard on the bench, which I happened to have laying around.

The end-point was that I was able to boot up from the replicate drive, 
and the system behaved as if nothing had changed. I was especially 
suprised to see that "format" reported the cloned maxtor drive to be a 
WDC, but possibly it is an indication of how "low level" the replication 
process actually is. (ie, the "30 gig maxtor" didn't realize it had 10 
gigs unallocated space any longer, thinking it was a "20 gig WDC" part.)

I would believe this procedure to be especially of use to anyone who 
might want to deploy a number of systems that are are very similar ; or 
possibly the paranoid folk who like to have a "spare system" on a disk, 
ready to drop into production in the event of hardware/other problems. 
Given that it is *much* faster than "filesystem level" cloning 
procedures I've used in the past (ie, ufsdump ; manually slice new disk 
; ufsrestore ; install boot-block) - and also quicker than Jumpstarting 
(I think?) ... it could be of interest to some others on the list, I hope.

Typical warning / caveat: Your mileage may vary ; the first time you try 
this, use a test system that you don't mind losing should trouble arise. 
(ie, in case you forget which is target vs donor . However, it was 
pretty brainless-straightforward to do.

I hope this information/summary doesn't offend those who feel the 
managers list is the exclusive domain of "high end" servers and topics 
servicing those exclusive needs.

--Tim Chipman

===for reference==

PCI IDE Raid card was "Sabrent RDIT" with an ITE IT8212f chipset. Google 
search for the terms "Sabrent RDIT" will locate the vendor I happened to 
use; clearly you are free to buy this or an equivalent part from 
wherever you might choose, if you wish to do so. Also, possibly of 
interest to some, the Raid Card appeared to recognize a 200gig Maxtor 
drive I have on the shelf without issue, suggesting support for "large 
ide drives" (>130-something gigs..) - although typically such large IDE 
drives don't actually get recognized at their full size by Solaris 
machines with IDE drive support, AFAIK.

For what interest it may be, I think this procedure works regardless of 
the OS that may happen to reside on the IDE disk in question. (Linux, 
Solaris, Windoze, etc).
sunmanagers mailing list
Received on Fri Mar 4 09:43:18 2005

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:43:44 EST