i'm trying to clone my CentOS 5 VM. Now I've read some tutorials on dd and it looked pretty straight forward.

I've booted my VM i'm cloning into rescue mode, mounted my NAS to place my image file, and used the following dd to clone the disk:

Now I created a new VM, booted it in rescue mode, mounted my NAS, I ran the following command to restore the image to this new VM:

its my understanding that when you dd the entire disk (/dev/sda) that this copies the MBR/partition table but when i run fdisk -l on the new VM, there is no partition table. When i reboot my new VM, there is no OS. Did I miss something? Did I miss a step? Do I need to backup/restore the MBR as well?

A VM has a single or set of files that can easily be copied in the HOST. Why don't you just copy the virtual hard drive and don't fool with the dd part in the client?

If I do dd on a real system I don't like to boot the original OS. I use a live cd/usb to access the real hard drive and put to some place.

Hi Jefro, yea I know, I've cloned my VM by copying the vmx files before. I'm just trying to learn how to use the dd command and how to clone a system.

I agree with you that this should work.

I just tried to copy the first 1k bytes from my /dev/sda to a file, and compared /dev/sda and the image file with hexedit. They are equal.
When you you open with khexedit the orginal /dev/sda, the .img file and the new /dev/sda, what do you see?

jlinkels

Then boot your VM to a live iso. Then dd it over. Be sure to use /dev/sda and not sda1.

hello jlinkels, I'm not famaliar with khexedit or hexedit. Any other ideas?

Well guyz, I think i figured it out, ran the dd command to copy the disk and I changed the follow bs option from 64K to 4096 and I also removed the notrunc as well. That seemed to fix the issue.

I used your commands many many times and they always worked, but I never used the conv=noerror,notrunc part of it.
Just a sidenote, if you are not speed limited (like a copy from disk to disk) you will get much better performance with larger blocksizes, I have had made good experiences with blocksizes between 4-16MB.

Reducing the BS down to 4096 awfully slows down the process. In addition, the conv=notrunc option was also used as an example in the learn the dd command thread, so I don't see a reason why that shouldn't work.

Maybe you could try to experiment with both settings (or neither) and check the result, just for the sake of learning.

jlinkels

I like to know about errors, so that is why I like dd and just a block size number.

Ok guyz, i just ran the same command on my VM in rescue mode

And then I restored it without any issues. I guess the 1st time around I tried it, there was some kind of disk/network hiccup.