On Fri, Apr 27, 2018 at 09:32:08PM -0700, Nick Leli wrote:
> The RAID is utilized with LVM, but I would like to inspect the
> individual disks, to ensure that data is replicated in case of
> failure.
You can run a RAID check and then see that mismatch_cnt is 0.
This should be enough to make sure all data is correct.
If you don't trust the RAID system at all, and provided all
devices are using the same data offsets, you can use cmp.
Setup:
# truncate -s 100M a.img b.img
# losetup --find --show a.img
/dev/loop0
# losetup --find --show b.img
/dev/loop1
# mdadm --create /dev/md42 --level=1 --raid-devices=2 /dev/loop[01]
mdadm: Defaulting to version 1.2 metadata
mdadm: array /dev/md42 started.
# mkfs.ext4 /dev/md42
# mkdir loop
# mount /dev/md42 loop/
# echo Hello World > loop/HelloWorld.txt
# umount /dev/md42
Test:
# cmp -l /dev/loop0 /dev/loop1
4257 0 1
4265 103 161
4266 210 64
4267 155 124
4268 313 203
4269 273 45
4270 154 216
4271 200 40
4272 351 206
4273 361 114
4274 264 262
4275 204 145
4276 21 204
4277 366 345
4278 140 144
4279 61 105
4280 126 265
4313 134 77
4314 53 372
4315 311 104
4316 264 333
So there are a few bytes difference, to be expected in metadata.
(device role 0 vs. device role 1, device UUID, checksum)
If you don't want to compare metadata, you'd have to map out a
loop device with losetup --offset / --sizelimit that represents
actual data on the RAID itself.
Checksums will only tell you that something is different,
not why and where (unless you do checksum per chunks of data).
Note this will only work for 1:1 mirrored RAID modes.
cmp would naturally complain about parity, or data offsets.
So for other RAID levels you have to rely on the built-in check
or use a tool that is aware of the specific RAID layout.
Regards
Andreas Klauer
--
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
