proper way to use testdisk to recover missing logical volumes on a raid 5 w/ LVM2
so i have 4 disks in a raid 5, /dev/sd[abcd]2......they were being recogized as /dev/md0 and i wanted them to be recognized as /dev/md2 to co-inside with the /dev/sd[abcd]2......so in a recovery mode from the ubuntu 9.04 server install cd, i removed /dev/md0 and created /dev/md2 as a raid 5 with 4 active disks....
But then my volume group (only had one, vg_system) was missing, and correspondingly all my logical volumes (lv_slash, lv_usr, lv_home, lv_tmp, lv_var, lv_opt)
I remember specifically that i had created lv_slash first, and w/ -L30G, so i re-created the pv and volume group (vg_system), and then added a 30G lv called lv_slash.....
So now i want to run testdisk to put the ext3 partition that was in lv_slash back so i can read the data, and provided that works, eventually figure out and recover the other 5 LVs on that volume group.
So my question is, what's the best way to run testdisk to recover these logical volumes? When i ran 'sudo testdisk /dev/mapper/vg_system-lv_slash' i was able to see the 30G partition, and could list the files that were there, but when i try and mount/fsck the LV, fsck complains about having a bad superblock and a non recognizable ext2 (which it's not, it's an ext3) superblock. It's almost like even though i'm running fsck.ext3, it's really running e2fsck for ext2 only....
ok, so i use testdisk to look for other superblocks, and write down all the block #'s, (4096 block size) and still no good w/ fsck.ext3, it complains about every single one of the -b <block> -B 4096 options i give it.
Should the disk geometry for testdisk when scanning LVs be something specific? I believe my CHS is <some huge number>,1,1