Removing and Replacing dm-intergrity Drives In A RAID Array     2026/08/06
Ami Sapphire

Last Updated: 2026/08/08


note the suspect drive and cross-check it with smartctl and your
integritysetup.sh entry list; they both list you serial or wwn

smartctl -i /dev/sdX

dmsetup info /dev/dm-X

--

unmount the array

umount /path/to/array/mount

note: you may need to stop the services that have used the drive - php,
mysql/mariadb, httpd, smbd, etc. - do not use umount flags -l or -f to remove
the array's mount!
the command to use if there are any processes using the array's mount

dmsetup ls

--

disable the array's automount in /etc/fstab by commenting out the entry, as
you do have to shut down the machine to replace the drive... unless you have a
hot spare installed

--

replace the suspect drive's entry with the replacement one in
integritysetup.sh - use either the serial number or wwn

--

apparently stopping the array makes the next step fail!
the command, for the record

mdadm --stop /dev/mdX

this would only be useful if the array had outright failed before replacement

--

fail and remove suspect drive

mdadm --manage /dev/mdX --fail /dev/dm-X --remove /dev/dm-X

--

replace the suspect drive with a known good one if you do not have a hot spare
installed - please cross-check the serial number or wwn of the suspect drive!

--

if the drive was just physically replaced, do the following (if an already
prepared spare, skip this step)

integritysetup format -B -I crc32c -s 512 -t 4 /dev/sdX

OR

integritysetup format -B -I crc32c -s 4096 -t 4 /dev/sdX

OR

integritysetup format -B -I sha1 -s 512 -t 20 /dev/sdX

OR

integritysetup format -B -I sha1 -s 4096 -t 20 /dev/sdX

then

integritysetup open -B -I crc32c /dev/sdX mdX-idY

OR

integritysetup open -B -I sha1 /dev/sdX mdX-idY

--

add the replacement drive

mdadm --manage /dev/mdX --add /dev/mapper/mdX-idY

if this fails with writing to the superblock, run your previous command, but
with the --integrity-recalculate flag - this will be very slow

Note: this should also only be done on a very reliable and known-good hard disk
and preferably on a system withh properly working ECC RAM

integritysetup open --integrity-recalculate -B -I crc32c /dev/sdX mdX-idY

OR

integritysetup open --integrity-recalculate -B -I sha1 /dev/sdX mdX-idY

and then attempt adding the drive to the array

strangely enough, a 'Device or resource busy' was emitted in my --integrity-
recalculate example, but then I tried to remove it, and... it was actually
added to the array and resyncing?! [Device md0-id0 is still in use.]

--

check array detail

mdadm --detail /dev/mdX

--

check overall array status

cat /proc/mdstat

--

live version of overall array status

watch cat /proc/mdstat

or

watch -n 1 /proc/mdstat

--

after the resync is finished, reboot to test

Note: if --integrity-recalculate was used on the replacement disk, this is gone
upon next reboot - it will be opened without the flag assuming you edited your
integritysetup.sh file correctly

--

run the following

mdadm --detail /dev/mdX

--

if everything is fine, uncomment the commented out entry in /etc/fstab and
reboot to test

--

you may want to add the following to /etc/mdadm.conf or /etc/mdadm/mdadm.conf
if not present

ARRAY /dev/mdX metadata=1.2 UUID=[uuid]

where [uuid] is your array's UUID

that UUID was listed when you checked the array's detail

--

run the following

update-initramfs -u