Extending an existing MD RAID raid0 array

# mdadm --add /dev/md127 /dev/loop2
mdadm: add new device failed for /dev/loop2 as 3: Invalid argument

# cat /proc/mdstat 
Personalities : [raid0] 
md127 : inactive sda[0] sdb[1] sdc[2](S)
      6282240 blocks super 1.2
       
unused devices: <none>

# man mdadm | sed -rn '/to convert a RAID0/,/^[[:blank:]]*$/{/^[[:blank:]]*$/q;p}' | fmt -w 80
       From 2.6.35, the Linux Kernel is able to convert a RAID0 in to
       a RAID4 or RAID5.  mdadm uses this functionality and the ability
       to add devices to a RAID4 to allow devices to be added to a RAID0.
       When  requested  to  do  this, mdadm will convert the RAID0 to a RAID4,
       add the necessary disks and make the reshape happen, and then convert
       the RAID4 back to RAID0.

# man mdadm | sed -rn '/If the --raid-disks option is being used to increase/,/^[[:blank:]]*$/{/^[[:blank:]]*$/q;p}' | fmt -w 80
              If the --raid-disks option is being used to increase the
              number of devices in an array, then --add can be used to add
              some extra devices to be included in the array.  In most cases
              this is not needed as  the  extra  devices  can  be  added as
              spares first, and then the number of raid disks can be changed.
              However, for RAID0 it is not possible to add spares.  So to
              increase the number of devices in a RAID0, it is necessary
              to set the new number of devices, and to add the new devices,
              in the same command.

The correct procedure to extend a raid0 array:

# cat /proc/mdstat 
Personalities : [raid0] [raid6] [raid5] [raid4] 
md127 : active raid0 loop1[1] loop0[0]
      10475520 blocks super 1.2 32k chunks
      
unused devices: <none>

The extension, or grow, itself - an example with new /dev/loop2 device (grow mode, level 'raid0', raid devices equal to '3' here, and new disk to add):

# mdadm -Gv /dev/md127 -l 0 -n 3 -a /dev/loop2

# cat /proc/mdstat 
Personalities : [raid6] [raid5] [raid4] [raid0] 
md127 : active raid4 loop2[3] loop1[1] loop0[0]
      10475520 blocks super 1.2 level 4, 32k chunk, algorithm 5 [4/3] [UU__]
      [>....................]  reshape =  2.1% (115168/5237760) finish=3.7min speed=23033K/sec
      
unused devices: <none>

# cat /proc/mdstat 
Personalities : [raid6] [raid5] [raid4] [raid0] 
md127 : active raid0 loop2[3] loop1[1] loop0[0]
      15713280 blocks super 1.2 32k chunks
      
unused devices: <none>

For an inactive raid0 array on SLES 12

If an existing raid0 array was attempted to be extend incorrectly (see 'Cause' section), the array (assuming the correct procedure hasn't been done before system reboot) would be in an inactive state. This is because the incorrect command line execution nevertheless caused metadata placement on the disk:

# cat /proc/mdstat 
Personalities : [raid0] 
md127 : inactive sda[0] sdb[1] sdc[2](S)
      6282240 blocks super 1.2
       
unused devices: <none>

# wipefs /dev/sdc
offset               type
----------------------------------------------------------------
0x1000               linux_raid_member   [raid]
                     LABEL: s124qb01:sf
                     UUID:  dcbae5a8-01aa-8599-d8bd-0cb3f2f6c436

See how '/dev/sdc' is visible as a disk member and marked as spare. This only affects old mdadm package versions.

Since raid0 conceptually does not have any spare device, the solution for seeing a spare device in raid0 caused by the older mdadm version is pretty simple:

# cat /proc/mdstat 
Personalities : [raid0] 
md127 : inactive sda[0] sdb[1] sdc[2](S)
      6282240 blocks super 1.2
       
unused devices: <none>

# mdadm /dev/md127 --remove /dev/sdc
mdadm: hot removed /dev/sdc from /dev/md127

# cat /proc/mdstat 
Personalities : [raid0] 
md127 : inactive sda[0] sdb[1]
      4188160 blocks super 1.2

# mdadm --run /dev/md127
mdadm: started array /dev/md/sf

# cat /proc/mdstat 
Personalities : [raid0] 
md127 : active raid0 sda[0] sdb[1]
      4188160 blocks super 1.2 256k chunks
      
unused devices: <none>

Do NOT forget to remove the MD superblock from the newly removed disk​:

# mdadm --zero-superblock /dev/sdc

Tunning array sync speed

# sysctl -w dev.raid.speed_limit_max=<value> # because reshaping causes huge i/o

Man pages

• https://manpages.opensuse.org/Tumbleweed/mdadm/mdadm.8.en.html