Page 1 of 1

RAID 5 parameters ext3

Posted: Tue Sep 20, 2016 9:43 pm
by rtdisk
I have a 7-row RAID 5 ext3 system with one missing/failed disk. The following RAID 5 parameters are known: 64k stripe size, right synchronous parity distribution, no parity delay, correct disk order.

Raid offset is *not* known (believed to be zero). [An ext3 example in the R-Studio manual like the NTFS example would be great]

So far poor/mixed results in seeing the original file/folder structure after scan - some files shown with correct names but marked as deleted (and corrupted); no undeleted files. However a fair number of large (very much larger then stripe size) files (.tar) found and restored using file type search.

Does the fact that large .tar files can be recovered correctly in the file type search confirm that the RAID configuration parameters are correct ?

Re: RAID 5 parameters ext3

Posted: Wed Sep 21, 2016 6:28 am
by RForce
Send me a PM if you'd like some remote assistance.

Re: RAID 5 parameters ext3

Posted: Thu Sep 22, 2016 6:46 am
by Alt
rtdisk wrote: Does the fact that large .tar files can be recovered correctly in the file type search confirm that the RAID configuration parameters are correct ?
What happened with the RAID?
Looks like the offset is incorrect.

Re: RAID 5 parameters ext3

Posted: Thu Sep 22, 2016 12:16 pm
by rtdisk
Thanks for your replies. Anything to improve my ignorance would be much appreciated !

Advice on how to confirm the RAID offset for ext3 RAID 5 would be very helpful. The entropy analysis programs did not produce a significant result.

To answer the RAID failure question, it was an older 3Ware 9650SE 8-LPML hardware RAID 5, 7 x 1 TB, no hot spares, under Linux, with ext3 file system. Disk #6 failed, then disk #3 gave ECC errors during rebuild. [For reasons not important here] everything except one directory was on the automated backup list. So extracted individual disks, numbered them correctly, and cloned each to image file. All cloned successfully except disk #6 which has a head stack failure.

I know RAID stripe is 64K. Manufacturer confirms right synchronous, no parity delay. Disk order is correct - matched by serial number against RAID info also. But don't know enough about how to determine RAID offset.

For RAID offset 0 or 1 sectors, and a missing disk inserted in slot #6, rstudio finds multiple 5.5 TB ext3 volumes (green).
a) They differ in "Parsed Boot Record" count from 2 to 4 to 6. Can I assume the one with 6 parsed MBRs is the right one ?
b) rstudio "Open Files" after scan on these partitions shows deleted files only, some file names correct, but very incomplete, and recovered files are corrupt.
c) Search by file type (raw files) with raid offset 0 can recover uncorrupted tar files much greater in size than Ndisk x stripe size. Does this confirm that RAID offset ?
d) Some volumes (green) cannot be viewed with "Open Files" - rstudio reports a corrupted file system (perhaps from interrupted RAID repair before failure). Should I write out these as 5.5TB disk images and try fsck or testdisk on them ?

Optimistically,
rtdisk

Re: RAID 5 parameters ext3

Posted: Fri Sep 23, 2016 8:18 am
by Alt
You may try to find the so-called magic number for ext file system: 53EF, it's on the 1080–1081 offset of the file system on the disk. Important: this is not an offset of the RAID, this is the offset of the file system. But I don't think that will help, most likely R-Studio has already found them as Recognized partitions.
BTW, do you include a missing disk as a replacement for Disk 6? And I also recommend you to check the RAID against its consistency to see where the its rebuild stopped: Checking RAID Consistency.