RAID basics with mdadm

RAID levels: what each trades.

# RAID 0: two 1 TB disks → 2 TB usable, zero redundancy
# One disk dies: all 2 TB lost

# RAID 1: two 1 TB disks → 1 TB usable, either disk can die
# Reads can come from either disk (read parallelism); writes go to both

# RAID 5: four 1 TB disks → 3 TB usable
# Parity is distributed: any one disk can fail
# Rebuild reads ALL remaining disks — risk window during rebuild

# RAID 6: four 1 TB disks → 2 TB usable
# Two parity stripes: survives any two simultaneous failures
# Slower writes than RAID 5 (double parity calculation)

# RAID 10: four 1 TB disks → 2 TB usable
# Mirrors first (1+1, 1+1), then stripes the mirror pairs
# Can survive 2 failures if they are in different mirror pairs

Creating a software RAID array with mdadm.

# Scenario: four 1 TB disks (/dev/sdb, /dev/sdc, /dev/sdd, /dev/sde)
# Goal: RAID 5 with 3 TB usable + 1 disk parity

# Install mdadm if not present
sudo apt install mdadm

# Create the RAID 5 array
sudo mdadm --create /dev/md0 \
  --level=5 \
  --raid-devices=4 \
  /dev/sdb /dev/sdc /dev/sdd /dev/sde
# mdadm: Defaulting to version 1.2 metadata
# mdadm: array /dev/md0 started.
# The array starts in a degraded state and begins syncing (building parity)

# Check sync progress (this takes hours for large arrays)
cat /proc/mdstat
# Personalities : [raid6] [raid5] [raid4]
# md0 : active raid5 sde[3] sdd[2] sdc[1] sdb[0]
#       3145728 blocks super 1.2 level 5, 512k chunk, algorithm 2 [4/4] [UUUU]
#       [===================>.]  resync = 96.8% (...)
#       finish=0.4min speed=144000K/sec

# Save the array config so it reassembles at boot
sudo mdadm --detail --scan | sudo tee -a /etc/mdadm/mdadm.conf
sudo update-initramfs -u   # include mdadm config in initramfs

# Format and mount once sync completes
sudo mkfs.ext4 /dev/md0
sudo mkdir /mnt/raid
sudo mount /dev/md0 /mnt/raid

# Add to /etc/fstab
# /dev/md0  /mnt/raid  ext4  defaults,nofail  0  2

Reading /proc/mdstat and mdadm —detail.

/proc/mdstat is your live RAID status window. Learn to read it quickly.

cat /proc/mdstat
# Personalities : [raid5]
# md0 : active raid5 sde[3] sdc[1] sdb[0]
#       3145728 blocks super 1.2 level 5, 512k chunk, algorithm 2 [4/3] [UUU_]
#
# unused devices: <none>

# Decode the status line:
# [4/3]   = array expects 4 disks, currently has 3 (one is missing/failed)
# [UUU_]  = disk state: U=up, _=missing/failed; 4th position is the failed disk
# sdd is absent from the device list — it failed

# For more detail:
sudo mdadm --detail /dev/md0
# ...
# Array Size : 3145728 (3.00 GiB 3.22 GB)
# Raid Level : raid5
# Raid Devices : 4
# Total Devices : 3
# State : clean, degraded
# Active Devices : 3
# Failed Devices : 1
# Spare Devices : 0
#
# Number   Major   Minor   RaidDevice State
#    0     8        16        0      active sync   /dev/sdb
#    1     8        32        1      active sync   /dev/sdc
#    -     0         0        2      removed        <-- failed slot
#    3     8        80        3      active sync   /dev/sde

LVM-RAID vs mdadm: when to use which.

Linux offers two paths to software RAID: mdadm (the classic kernel md layer) and LVM’s built-in RAID support (introduced in LVM 2.02.89).

# LVM RAID: create a mirrored LV directly in LVM
sudo lvcreate --type raid1 -m1 -L 100G -n dbdata myvg
# -m1 = 1 mirror (2 copies total, like RAID 1)
# LVM manages the RAID internally — no /dev/md* device

# Check LVM RAID status
sudo lvs -a -o +devices,raid_sync_action myvg
#   LV            Attr       LSize  Devices                    Sync Action
#   dbdata        rwi-aor--- 100.00g dbdata_rimage_0(0),dbdata_rimage_1(0) idle
#   [dbdata_rimage_0] Iwi-aor--- 100.00g /dev/sdb(0)
#   [dbdata_rimage_1] Iwi-aor--- 100.00g /dev/sdc(0)

For new setups where you are already using LVM, LVM-RAID is cleaner. For arrays shared across servers (e.g. iSCSI targets) or where you want maximum documentation coverage, mdadm remains the standard.

RAID is not a backup. Understand the failure modes it does not cover.

# What RAID protects against:
# - A single physical disk dying (RAID 1/5/6/10)
# - Two disks dying simultaneously (RAID 6, RAID 10 in some configs)

# What RAID does NOT protect against:
# - Accidental rm -rf /mnt/raid/important-dir
#   -> deleted from all mirrors instantly
# - Ransomware encrypting files
#   -> encrypted on all disks simultaneously
# - Filesystem corruption (kernel bug, power loss mid-write, bad RAM)
#   -> written to all mirrors; the corruption is mirrored
# - RAID controller failure (all disks attached to a dead controller)
#   -> array unusable until controller replaced or disks moved
# - Silent data corruption (bit rot)
#   -> RAID 5/6 parity can detect but not always correct silent flips;
#      use btrfs or ZFS with checksums for full bit-rot protection

# The mental model:
# RAID = high availability (tolerate hardware failure)
# Backup = data protection (recover from data loss)
# Both are required. Neither replaces the other.