# For the sake of briefness, we use Bash  "group compound" stanza:
#   /dev/sd{a,b,...}1 => /dev/sda1 /dev/sdb1 ...
# Along the following variables:
#   ${M} array identifier (/dev/md${M})
#   ${D} device identifier (/dev/sd${D})
#   ${P} partition identifier (/dev/sd${D}${P})

# Create (initialize) a new array
mdadm --create /dev/md${M} --level=raid5 --raid-devices=4 /dev/sd{a,b,c,d,e}${P} --spare-devices=/dev/sdf1

# Manually assemble (activate) an existing array
mdadm --assemble /dev/md${M} /dev/sd{a,b,c,d,e}${P}

# Automatically assemble (activate) all existing arrays
mdadm --assemble --scan

# Stop an assembled (active) array
mdadm --stop /dev/md${M}

# See array configuration
mdadm --query /dev/md${M}

# See array component configuration (dump superblock content)
mdadm --query --examine /dev/sd${D}${P}

# See detailed array confiration/status
mdadm --detail /dev/md${M}

# Save existing arrays configuration
# (MAY be required by initrd for successfull boot)
mdadm --detail --scan > /etc/mdadm/mdadm.conf

# Erase array component superblock
# (MUST do before reusing a partition for other purposes)
mdadm --zero-superblock /dev/sd${D}${P}

# Manually mark a component as failed
# (SHOULD when a device shows wear-and-tear signs, e.g. through SMART)
mdadm --manage /dev/md${M} --fail /dev/sd${D}${P}

# Remove a failed component
# (SHOULD before preemptively replacing a device, after failing it)
mdadm --manage /dev/md${M} --remove /dev/sd${D}${P}

# Prepare (format) a new device to replace a failed one
sfdisk -d /dev/sd${D,sane} | sfdisk /dev/sd${D,new}

# Add new component to an existing array
# (this will trigger the rebuild)
mdadm --manage /dev/md${M} --add /dev/sd${D,new}${P}

# See assembled (active) arrays status
cat /proc/mdstat

# Rename a device
# (SHOULD after hostname change; eg. name="$(hostname -s)")
mdadm --assemble /dev/md${M} /dev/sd{a,b,c,d,e}${P} --name="${name}:${M}" --update=name