On 04/01/16 15:14, Artur Paszkiewicz wrote: > It looks like the metadata on the drives got messed up for some reason. > If you believe the drives are good, you can try recreating the arrays > with the same layout to write fresh metadata to the drives, without > overwriting the actual data. In this case it can be done like this (make > a backup of the drives using dd before trying it): > > # mdadm -Ss > # mdadm -C /dev/md/imsm0 -eimsm -n4 /dev/sdd /dev/sda /dev/sdc /dev/sdb -R > # mdadm -C /dev/md/md0 -l5 -n4 /dev/sdd /dev/sda /dev/sdc /dev/sdb --size=900G > --chunk=128 --assume-clean -R > # mdadm -C /dev/md/md1 -l5 -n4 /dev/sdd /dev/sda /dev/sdc /dev/sdb --chunk=128 > --assume-clean -R > > Drives should be listed in the order as they appear in the output from > mdadm -E. Look at the "DiskXX Serial" lines. > > Then you can run fsck on the filesystems. Finally, repair any mismatched > parity blocks: That sounds like it'll write to the superblock. Okay - I guess needs must. BUT!!! Once you've done that, create the array using loopback/overlays or whatever (I hope someone else chimes in here :-) Whatever, the disks themselves will be read-only, so you can run fsck, check that your re-assemble really did work properly - if it did fsck won't find much wrong - and if it worked right then you can do it for real. Thing is, this means that if you do go wrong, you haven't actually done anything to the data on the disk itself, so you can throw the overlay away and try afresh, without having to restore your backup. Cheers, Wol -- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
