Cloning an entire drive?

[spiderpig]

I've had this question for some time, but have never asked...

There is an article from the OpenBSD Journal which covered cloning drives:

http://undeadly.org/cgi?action=artic...20080319204706

The author advocates using dump & restore on each partition, but if cloning the entire drive is the goal, why not simply back up the c partition?

[Eam404]

Well if by C drive you mean /root partition then no you cant just backup your root partition. Things like usrland apps and /etc configuration files are all items that most people like to back up; at lest generally. Dump/Restore works great on BSD systems. If you are wanting a more universal way to "clone" your hard drive the dd command works wonders. Heres a small example: http://www.linuxweblog.com/dd-image

[spiderpig]

Originally Posted by Eam404

Well if by C drive you mean /root partition then no you cant just backup your root partition.

Actually, my terminology was quite precise. I assume what you refer to /root is typically /dev/wd0a on a bootable IDE drive, or /dev/sd0a is the bootable drive is SCSI. Otherwise, /root is merely a directory, not a partition.

Again, given that /dev/wd0c or /dev/sd0c both covers the entire disk and is accessible, why not clone it instead of each individual partition?

[revzalot]

Thanks for bringing this up and I've always wanted to find a way to clone drives. One way is using g4u but I want to clone a drive by using the same box. Is it possible to just go into single user mode, mount the two drives:

# mount -a /dev/wd0
# mount -a /dev/wd1

then use the copy command:

# cp /dev/wd0 /dev/wd1


Is this possible assuming both drives are the same capacity.

[spiderpig]

Originally Posted by revzalot

Is this possible assuming both drives are the same capacity.

No, for at least two reasons. One, cp(1) copies relative to the user that invoked the command. If cp is invoked as a user other than root, then permissions may be set which will not allow that user to read and copy. If root is used to copy files, then individual user ownership is lost. Two, consider all the device nodes residing in /dev. The script /dev/MAKEDEV is executed at installation time for a reason.

[ocicat]

Originally Posted by spiderpig

No, for at least two reasons.

If the drive being cloned is a boot drive, you should also consider running installboot(8) as the article suggests.

[BSDfan666]

For one revzalot, /dev/{wd0,wd1} don't exist.. and you can't use "cp" like that either.

To backup a partition, I use either tar or dump, if I'm trying to recover something.. like lost data off a friends thumb drive or w/e, I duplicate the partition with dd...

$ dd if=/dev/rsd0i of=partition-image.img
Then work on the image instead of risking the drive..

[J65nko]

Originally Posted by spiderpig

The author advocates using dump & restore on each partition, but if cloning the entire drive is the goal, why not simply back up the c partition?

dump and restore act on filesystems.From the dump man page

NAME
     dump - filesystem backup

Label "c" does not have a filesystem, it is just a virtual label encompassing the complete disk, which could have other operating systems installed.

From one of my systems

$ disklabel wd0
# Inside MBR partition 0: type A6 start 63 size 81915372
# /dev/rwd0c:
type: ESDI
disk: ESDI/IDE disk
label: WDC WD3200AAKS-2
flags:
bytes/sector: 512
sectors/track: 63
tracks/cylinder: 255
sectors/cylinder: 16065
cylinders: 38913
total sectors: 625142448
rpm: 3600
interleave: 1
trackskew: 0
cylinderskew: 0
headswitch: 0           # microseconds
track-to-track seek: 0  # microseconds
drivedata: 0 

16 partitions:
#                size           offset  fstype [fsize bsize  cpg]
  a:         81787372               63  4.2BSD   2048 16384    1 
  b:           128000         81787435    swap                   
  c:        625142448                0  unused      0     0      
  i:        204796620        163830870 unknown                   
  j:        256509855        368627490   MSDOS

See, no fstype defined for label "c"

I have some other OS's installed

]fdisk wd0
Disk: wd0       geometry: 38913/255/63 [625142448 Sectors]
Offset: 0       Signature: 0xAA55
          Starting         Ending        LBA Info:
 #: id      C   H  S -      C   H  S [       start:        size ]
------------------------------------------------------------------------
*0: A6      0   1  1 -   5098 254 63 [          63:    81915372 ] OpenBSD     
 1: A9   5099   0  1 -  10197 254 63 [    81915435:    81915435 ] NetBSD      
 2: A5  10198   0  1 -  22945 254 63 [   163830870:   204796620 ] FreeBSD     
 3: 05  22946   0  1 -  38912 254 63 [   368627490:   256509855 ] Extended DOS

These all lie within the boundaries of the label "c" size of 625,142,448 sectors.
dump is not so smart it can backup NetBSD, FreeBSD and the extended DOS partitions in one go

[BSDfan666]

What J65nko posted *should* be common knowledge, I'm not even going to comment on Eam404's reply.

[Eam404]

Welp, I guess thats what I get for not reading the whole post...

[neant]

I hope there's no problem necroing this thread since it's a sticky.

For the record, cloning an entire disk through the c partition can be done, using dd(1), HOWEVER, this can be dangerous. There's this post from 2001* by Randy Lewis, it basically says:

Example:
                                       +----------- bytes/sector value
                                      /    +------- sectors/cylinder value
                                     /    /
% dd if=/dev/rsd0c of=/dev/rsd1c bs=512x4520 count=7931 <cr>
                                                    /
                 total cyliders to copy value -----+

First, the obvious, dd is dumb, it shouldn't be run on a mounted/live disk. Incomplete files, inconsistent fs and partitions marked unclean since they were never unmounted are to be expected. Which means the clone will have to be fscked before use.
Second, OpenBSD 5.0+ (AFAIK) uses disk UIDs, fstab by default doesn't have /dev/sd0 in it, it has the ID of the disk instead. The clone, since it's an exact copy, will end up having the same DUID as the original, not good if the clone isn't pulled out of the system right away. The ID can be changed using disklabel(8)'s "i" command in the editor.

On the other hand, assuming two identical disks, reboot the system from a live cd (single-user works too I guess), run dd, pull out the clone and keep it in storage, and you've got a full system backup ready to go in as long as it takes to change a disk.
Didn't actually try this last part yet, but there's no reason it shouldn't work, right?

* Can't post links. The post can be found at marc.info /?l=openbsd-tech&m=100765187914137&w=2, or just search for "Re: Cloning OpenBSD disks (amended / expanded explanation)" on google.

[joancatala]

Some days ago I wrote a brief article on my personal blog about cloning easily disks (whether they are SATA or SSD). The article is in catalan: http://joancatala.net/node/1419

Basically, I burned an Ubuntu (live) image.
I did dd if=MY_DISK of=BACKUP_COPY_IMAGE bs=32M
I changed the disk, from SATA3 to SSD.
Finally I did if=BACKUP_COPY_IMAGE of=MY_NEW_DISK bs=32M

Simple solution.
Buy some beers, because is a lot of time (about 4/5 hours).

[bsd-keith]

My 'cloning' program has been dd for quite some time, on both BSDs & Linux distros.

Fairly easy to learn to use, & found on most, if not all, unix like systems, just double check your input & output before committing.