Hello,
I have a Raid5 array where the controller card took a dump on me.
I am reasonably sure the drives are in good working condition... and i am looking for a way to get my data back.
The Raid5 array consists of 8 x 1.5 TB SATA drives... the fun part is that the FS is ZFS.
I love ZFS... but since there are no utilities that can rebuild a raid array and read ZFS... I have to get creative.
(Also before I get grief for not using RAIDZ, the only controller I had was an older LSI controller that did not support JBOD... I would not have this issue if I would have used RAIDZ instead)
Getting back to the issue...
I know I will not be able to use R-Studio to fix the array, but I am hoping to recover my files. The seach for files feature seems to work as it does find a large amoun of my small files (jpgs that did not span a drive).
I read the recovery guide, and the finding raid parameters guys, but honestly, they are hard to follow... especially if you do not have an NTFS FS to follow along with.
Here is what I think I know...
I know there are either 8 drives or 6 drives (more about this below)
I know it is a Raid 5 Array
I think the block size is 128K (I was able to recover that from the raid controller config... but i am only about 90% sure it is right)
Here is what I don't know...
I don't know the Disk Order
I don't know the Raid Offset
I don't know the Block Order
I followed the raid parameter guide as best as I could and I gathered the following informaiton using the MBR search technique:
Drive ID Location of MBR's first byte
PD0 2c186a55
PD1 2de00b38
PD2 7430ef21
PD3 2de00b38
PD4 2c1b9255
PD5 2c1c3201
PD6 2c189a01
PD7 a9bdf055
It seems that two of the drives have a huge variance in comparison to the other six drives... i don't know if that makes any difference. There is a possibility that PD1 failed and was pulled from the array (it's SMART statistics do not look good).
There is a possibility that PD2 and PD7 were not acutally members of the array, I cannot recall how much space the full array had... and this was setup a long time ago.
I don't know where go next. I used escel to sort the data above and got the following results (which I thought might be my disk order)
Drive ID Location of MBR's first byte
PD0 2c186a55
PD6 2c189a01
PD4 2c1b9255
PD5 2c1c3201
PD1 2de00b38
PD3 2de00b38
PD2 7430ef21
PD7 a9bdf055
I tried that and it did not seem to work (I could read most jpgs that were smaller than 128K, but nothing larger).
Any help would be appreciated. I am ready and willing to purchase R-Studio if it can help me recover some of my lost data. There are years of family files on there that I would hate to lose. I am in the process of building a new SAN/NAS at home and this one bit the dust just a little too soon (Murpy's Law )
Need help putting my Raid5 back together
Re: Need help putting my Raid5 back together
There's a free utility to find RAID parameters: ReclaiMe Free RAID Recovery. As far as I understand, it works regardless of the file system used in the RAID. Maybe, it can help. Otherwise, I simply don't know what to do. Exotic file systems have their disadvantages, too.
Re: Need help putting my Raid5 back together
Alt,
Thank you for the reply.
I was evaluating the Free RAID Recovery tool you mentioned when I read your post... I think it will be of great help in this situation.
It looks like there were some other issues going on with the Array that will make this a bit harder, I have a few questions about R-Studio...
Can the full version of R-Studio mount the Virtual raid array? I am using the Windows version right now and that would be an invaluable option (I don't have the space to create a 12 TB image file so creating a single image and mounting that is not currently an option).
Here is my plan so far:
1. Use the ReclaiMe Free RAID Recovery utility to map the RAID array and get the proper parameters (Done)
2. Use R-Studio to create the Virtual RAID Array and mount it to the OS (assuming R-Studio can do this, if not I have to try to make a huge image file somewhere and mount that instead).
3. Due to the Parittion being ZFS, I cannot use ReclaiMe or R-Studio to recover the data. This is especially true since I was using DeDuplication inside the Z-Pool so a Scan for Files will not work properly as some blocks of data are pointers to the DeDupe block (this is why adding native ZFS support to R-Studio would make it the one RAID recovery tool to rule them all). I will have to use TestDisk (which has full ZFS support) to recover files and folders from the mounted Array/Image (Which means I will need at least another 5TB of space for the recovery operation).
After that I will full backup of the recoverable files... I can reclaim all the temporary space and move on knowing I got back what I could.
I have seen a lot of requests for ZFS recovery help (here and other places) since it is such an advanced FS there are not alot of external tools to work on it. ZFS may be an exotic file system, but it will become very mainstream for NAS devices in the near future. Since Oracle had to "donate" it to the Open Source community (in an attempt to avoid littigation from NetAPP)* and Open Solaris is no longer going to have Security updates provided to users the new home of ZFS is in the FreeBSD platform. FreeNAS is taking the bull by the horns and XiSystems is really doing a great job keeping FreeNAS updated. Adding full ZFS support to R-Studio would be a really good idea.
* NetAPP threatend litigation against Oracle because ZFS is far too similar to their proprietary, pattented, Data On-Tap OS (RAID-DP architecture with Write Anywhere File Layer aka WAFL). Oracle didn't have a chance to win so they "donated" it to the Open Source community.
Anyhow, it will be a long time before I can drum up enough storage to make this happen (unless R-Studio can mount the Virtual Array using the acutal disks) I will post my results once I have everything I need to run with my plan.
Thank you for the reply.
I was evaluating the Free RAID Recovery tool you mentioned when I read your post... I think it will be of great help in this situation.
It looks like there were some other issues going on with the Array that will make this a bit harder, I have a few questions about R-Studio...
Can the full version of R-Studio mount the Virtual raid array? I am using the Windows version right now and that would be an invaluable option (I don't have the space to create a 12 TB image file so creating a single image and mounting that is not currently an option).
Here is my plan so far:
1. Use the ReclaiMe Free RAID Recovery utility to map the RAID array and get the proper parameters (Done)
2. Use R-Studio to create the Virtual RAID Array and mount it to the OS (assuming R-Studio can do this, if not I have to try to make a huge image file somewhere and mount that instead).
3. Due to the Parittion being ZFS, I cannot use ReclaiMe or R-Studio to recover the data. This is especially true since I was using DeDuplication inside the Z-Pool so a Scan for Files will not work properly as some blocks of data are pointers to the DeDupe block (this is why adding native ZFS support to R-Studio would make it the one RAID recovery tool to rule them all). I will have to use TestDisk (which has full ZFS support) to recover files and folders from the mounted Array/Image (Which means I will need at least another 5TB of space for the recovery operation).
After that I will full backup of the recoverable files... I can reclaim all the temporary space and move on knowing I got back what I could.
I have seen a lot of requests for ZFS recovery help (here and other places) since it is such an advanced FS there are not alot of external tools to work on it. ZFS may be an exotic file system, but it will become very mainstream for NAS devices in the near future. Since Oracle had to "donate" it to the Open Source community (in an attempt to avoid littigation from NetAPP)* and Open Solaris is no longer going to have Security updates provided to users the new home of ZFS is in the FreeBSD platform. FreeNAS is taking the bull by the horns and XiSystems is really doing a great job keeping FreeNAS updated. Adding full ZFS support to R-Studio would be a really good idea.
* NetAPP threatend litigation against Oracle because ZFS is far too similar to their proprietary, pattented, Data On-Tap OS (RAID-DP architecture with Write Anywhere File Layer aka WAFL). Oracle didn't have a chance to win so they "donated" it to the Open Source community.
Anyhow, it will be a long time before I can drum up enough storage to make this happen (unless R-Studio can mount the Virtual Array using the acutal disks) I will post my results once I have everything I need to run with my plan.
Re: Need help putting my Raid5 back together
Unfortunately, R-Studio cannot pass virtual RAIDs, those objects exist only within R-Studio, so you'll have to image the RAID to some other place.