I am working on consistent snapshots verifications. We snapshotted the raw device attached (saying two drives sdb and sdc) to a host and want to check the data was preserved with the desired orders as they had been write to the disk, we want to prove the difference in the data snapping time sequence between the normal snaps and the consistent snaps, the normal snaps meant to snap two drives separately, and the consistent snaps mean to snap them together in one consistency group.
As state above, I create a EXT4 file system on the LVM spawning in three drives in Raid0 with following commands.
pvcreate /dev/vdaa /dev/vdab /dev/vdac
vgcreate test /dev/vdaa /dev/vdab /dev/vdac
lvcreate --type raid0 -L 2G --stripes 3 --stripesize 4 -n lv test
mkfs.ext4 /dev/test/lv
mount /dev/test/lv /mnt
Then I tried to:
Snapshot vdaa, vdab and vdac separately one by one with normal snapshots, and them rollback them and check the data integrity, there is data loss in the file system, but the file system can be pulled up.
Take a consistent snaps to all the three devices and rollback them, there is no data integrity issue,
However, I am wondering how the data is laying out in the case 1? and why the file system could be pulled up after rollbacking? I thought the file system would stay offline as the data was not consistent in time because their data was snapped separately in 3 time points (even if close enough)
