The way I'm using it in my script (the image is a PARTITION, not a HD), I shouldn't have to say -o, cause that works for virtual DISKS.

http://www.clarkson.edu/projects/itl...ddtutorial.txt

Don't you agree?

Update:

Well, It's been a long time.

I have made some advances... though not the way I wanted. I wanted to let linux do the hard work for me (bring the raid to life itself), but haven't been able to do that. So what I DID do was make a program that rebuilt the binary content of the drive that the raid logically contains.

I made a program that takes the separate images and calculates the layout of the resulting drive. I make an Image and then I follow this tutorial to mount its content: http://www.clarkson.edu/projects/itl...ddtutorial.txt

It takes like two hours to generate the image (72 GB aprox). I made a run using Left Asymmetrich, but failed. I could ls the root directory and see some of the entries in there, but that was about it. So I'm running symmetrich right now. Let's see If I make it.

Yes, you do have to use -o when using image disks. As I say this again, the partition starts some where. If you are using 16 heads, 63 sectors per track, and 512 bytes per sector, the partition table information starts at 32256. If you are not using this geometry, you will have to calculate yourself. I suggest using 516096c or 516096 bytes per cylinder for the bs option for dd instead of 1M. This makes it easy to calculate the location where the patition table starts at.

But Electro.... the problem is that in the images that make up a raid, there's no Partition Table. Am I wrong? The only parttition table in those images would have to be in the logical content (the virtual content) of the raid, right?

Well.... thinking about it a little better, what you mean is that I should actually use a partition created inside the image.... not the image itself. Something like using /dev/hda1 instead of /dev/hda, right? I'll have to give it a try.

It's been a rather long journey.... but I finally got it!

Just like a told you. It wasn't solved the way I expected to... but I certainly solved it.

Using the program I made (and told you about) I used ther four images to rebuild the logical content of the raid in a single file (I and some coworkers had to check each image to see how they were ordered in the array and the algorithm used by the contoller). Then I used that file as the image of a HD. Checked its Partition Table, and mounted accordingly (using losetup).

I want to thank all you, guys, that helped me with this task. I wouldn't have been able to get it accomplished without your help!

What I recovered: 60 GBs of data.

Guys, I made a request to make an article based on this experience... and it has already been published by the Free Software Magazine. I hope you enjoy it! Thank you very much, guys!

http://www.freesoftwaremagazine.com/...recovery_raid/