On 30/07/19 12:07, Gandalf Corvotempesta wrote: > # Sync disk partitions > sfdisk --dump /dev/sda | sfdisk /dev/sdc > sfdisk --dump /dev/sdb | sfdisk /dev/sdd > > # Rebuild array > mdadm /dev/md0 --add /dev/sdc1 > mdadm /dev/md0 --replace /dev/sda1 --with /dev/sdc1 > mdadm /dev/md0 --add /dev/sdd1 > mdadm /dev/md0 --replace /dev/sdb1 --with /dev/sdd1 > > This should replace, with no loss of redundancy, sda with sdc and sdb with sdd. > Then I have to re-install the bootloaded on new disks and reboot to > run from the SSDs > > Any thoughts ? What about LVM ? By syncing the disk partitions and > the undelying array, LVM should be up & running on next reboot > automatically, even if moved from SAS to SSD, right ? Something has come up on the list recently, called dm-integrity. It's worth thinking about. Once you've created your sdc1 and sdd1, consider creating a dm-integrity device on them, and then adding that dm-integrity device into your mirror. It's another layer in the stack, but it check-sums and ensures read integrity. So should something *corrupt* one of your mirrors, the dm-integrity layer will fail with a read error, and the raid will correct the mirror for you. The alternative is a raid that returns correct and corrupt data at random, until you do an integrity check, which will toss a dice whether it overwrites the corrupt data with the correct data, or vice versa. Please note that dm-integrity is NEW, so use it at your own risk ... Cheers, Wol
