Thank you John. The thing is my data was not overwritten, corrupted, etc. Some was, but I know which parts. Basically, I just cleared the file system designation. So if a file is 64K, does the first 32K on drive 1 contain the first half of the file and the first 32K on drive 2 contain the second half, or are the 32 size chunks on random locations?

Thanks, Jeff

On Fri, Mar 5, 2010 at 6:18 PM, John R Pierce pierce@hogranch.com wrote:

Jeff Sadino wrote:

...

I have access to all of the HEX data on each drive, and I now which sector each stripe starts at. Is there any way that I can reconstruct my data from that? When a file gets split up in RAID0, does the controller use the same sectors on each stripe to write the file parts?

if a raid0 has a stripe size of 32k, then every other 32k is written on one drive, and every other 32k is written on the next drive.

if you lose one drive, then every other 32k of your file system is wiped out, gone, kaput. its like taking a 12-gauge shotgun to your data and directory information.

---

CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos

On Mar 5, 2010, at 11:36 PM, Ross Walker rswwalker@gmail.com wrote:

On Mar 5, 2010, at 11:22 PM, Jeff Sadino jsadino.queens@gmail.com wrote:

...

Thank you John. The thing is my data was not overwritten, corrupted, etc. Some was, but I know which parts. Basically, I just cleared the file system designation. So if a file is 64K, does the first 32K on drive 1 contain the first half of the file and the first 32K on drive 2 contain the second half, or are the 32 size chunks on random locations?

Trying to restich a broken RAID0 is going to be difficult. Question is is the md raid metadata ok, if it isn't then you are hosed. Make sure the metadata, usually stored in the last 64k+ of the partition is ok then see if you can reassemble the array, then try using a disk or file system rescue utility on a copy of the md raid block device and see what data you can recover.

Let me expand by saying.

1 make sure the partition table is restored to it's exact prior setup.

2 verify the md raid metadata at the end of the raid partition is ok. Need to research that one.

3 reassemble the raid0

4 make a copy of the raid device

5 use recovery tools on the copy to retrieve as much data as you can

-Ross

Hello Everyone,

Thanks to everyone who helped me to learn about RAIDs, starting at the end and working backwards :)

I brought my system back up, minus the raid0 partition that was corrupted. My question now is what's the best way to back up my server? Basically, I have one more (software?) raid1 array, a cluster setup on 5 subnodes (but I don't see any of their data here, so prob just all scratch and OS space), and then a hardware RAID controller (10.54.1.100) attached to 2 raid setups (scratch, apps, and data as one setup and then last week we added the shacks as a seperate 12TB array. It came as a RAID, but I think we had to decouple it b/c our OS couldn't read it), and then we have the computer with mounts referencing our main Windows server (10.1.1.17), which is backed up by our Windows guy. Should I use the empty shacks as backup space?

I hope that all makes sense. If not, please let me know. I really want to back up and ensure thsi never happens again. Thank you!

This is the output from `df -h`: Filesystem Size Used Avail Use% Mounted on /dev/sda1 9.8G 7.5G 2.3G 77% / none 4.0G 0 4.0G 0% /dev/shm /dev/md0 459G 140G 296G 33% /export /dev/sda2 2.0G 1.4G 561M 72% /var tmpfs 241M 3.8M 238M 2% /var/lib/ganglia/rrds 10.54.1.100:/mnt/RAID/FSDATA 1.1T 812G 223G 79% /data 10.54.1.100:/mnt/RAID/apps 1.1T 812G 223G 79% /apps 10.54.1.100:/scratch 72G 13G 55G 19% /scratch 10.54.1.100:/mnt/shack1 2.0T 103M 1.9T 1% /mnt/shack1 10.54.1.100:/mnt/shack2 1.8T 238G 1.5T 14% /mnt/fs3 10.54.1.100:/mnt/shack3 1.8T 285G 1.5T 17% /mnt/fs4 10.54.1.100:/mnt/shack4 1.8T 100M 1.7T 1% /mnt/shack4 10.54.1.100:/mnt/shack5 1.8T 100M 1.7T 1% /mnt/shack5 10.54.1.100:/mnt/shack6 1.7T 101M 1.6T 1% /mnt/shack6 //10.1.1.17/Scanner_data2 11T 4.7T 5.4T 47% /mnt/scanner_data2 //10.1.1.17/SCANNER_DATA 11T 4.7T 5.4T 47% /mnt/scanner_data //10.1.1.17/shared 11T 4.7T 5.4T 47% /mnt/shared //10.1.1.17/fMRI 11T 4.7T 5.4T 47% /mnt/fMRI //10.1.1.17/USERS 11T 4.7T 5.4T 47% /mnt/users /export/home/fs431 459G 140G 296G 33% /home/fs431 /export/home/coreg 459G 140G 296G 33% /home/coreg

On Fri, Mar 5, 2010 at 7:26 PM, John R Pierce pierce@hogranch.com wrote:

Jeff Sadino wrote:

...

Thank you John. The thing is my data was not overwritten, corrupted, etc. Some was, but I know which parts. Basically, I just cleared the file system designation. So if a file is 64K, does the first 32K on drive 1 contain the first half of the file and the first 32K on drive 2 contain the second half, or are the 32 size chunks on random locations?

didn't you say you did a mkfs ext3 over this stripe's partition? that would have overwirtten all the root directory areas, made a total mess of things.

ifyou -just- used a fdisk program to change the partition type, quick, change it back to what it was, and put it all back.

---

CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos