RAID (Redundant Array of Independent Disks) is a powerful way to boost performance, increase fault tolerance, or both. While many desktop users rely on GUI tools, seasoned sysadmins often prefer the control and transparency of a command‑line implementation. In this guide we’ll walk through installing and configuring a software RAID array on a modern Linux distribution, covering everything from hardware prep to verification and troubleshooting. By the end you’ll have a fully functional RAID 1 (mirrored) or RAID 5 (striped with parity) setup that you can tailor to your workload.
What You’ll Need
- At least two (RAID 1) or three (RAID 5) identical or similar hard drives or SSDs.
- A Linux system with root (or sudo) access – Ubuntu 22.04, Debian 12, CentOS 9 Stream, or any distro that ships
mdadm. - Basic familiarity with the terminal and partitioning tools (fdisk, gdisk, or partprobe).
- Backup of any data on the drives you plan to use – RAID creation wipes them.
- Internet connection to install missing packages.
Step 1: Install the mdadm Utility
The mdadm package is the de‑facto standard for managing Linux software RAID. Install it with your distro’s package manager:
# Debian/Ubuntu
sudo apt update && sudo apt install mdadm -y
# RHEL/CentOS/Fedora
sudo dnf install mdadm -y
During installation you may be prompted to configure mdadm.conf. Choose the default (auto‑generate) unless you have a custom initramfs requirement.
Step 2: Identify and Prepare the Disks
List all block devices and note the identifiers for the disks you’ll use (e.g., /dev/sdb, /dev/sdc, /dev/sdd).
# List devices
lsblk -d -o NAME,SIZE,MODEL,TYPE
For each disk, wipe any existing partition table to avoid confusion:
# Replace /dev/sdx with your actual device
sudo wipefs -a /dev/sdx
sudo dd if=/dev/zero of=/dev/sdx bs=1M count=10 status=progress
Now create a single Linux RAID partition on each drive. You can use fdisk (MBR) or gdisk (GPT). Here we use gdisk for GPT compatibility:
sudo gdisk /dev/sdx
# Inside gdisk:
# o – create a new GPT (if not already)
# n – new partition, default start, default end, type FD00 (Linux RAID)
# w – write and exit
Repeat for every drive you intend to include in the array.
Step 3: Create the RAID Array
With the partitions ready (e.g., /dev/sdb1, /dev/sdc1, /dev/sdd1), invoke mdadm to build the array. The syntax differs by RAID level:
# RAID 1 (mirroring) – two disks
sudo mdadm --create /dev/md0 --level=1 --raid-devices=2 /dev/sdb1 /dev/sdc1
# RAID 5 (striped with parity) – three or more disks
sudo mdadm --create /dev/md0 --level=5 --raid-devices=3 /dev/sdb1 /dev/sdc1 /dev/sdd1
After issuing the command, mdadm will display a progress summary. You can monitor synchronization with:
cat /proc/mdstat
Depending on drive size, initial sync may take several minutes to hours. Let it finish before proceeding.
Step 4: Create a Filesystem on the RAID Device
Once the array is healthy (clean state in /proc/mdstat), format it with the filesystem of your choice. Ext4 is a solid default; XFS or Btrfs are alternatives for larger or more feature‑rich setups.
# Ext4 example
sudo mkfs.ext4 -L raiddata /dev/md0
# XFS example
# sudo mkfs.xfs -L raiddata /dev/md0
The -L flag assigns a label, making future mounts easier.
Step 5: Mount the RAID Array and Make It Persistent
Create a mount point and mount the new filesystem:
sudo mkdir -p /mnt/raid
sudo mount /dev/md0 /mnt/raid
Verify the mount:
df -hT | grep raid
To ensure the array mounts automatically after reboot, add an entry to /etc/fstab. Use the label for resilience against device name changes:
echo 'LABEL=raiddata /mnt/raid ext4 defaults,nofail 0 2' | sudo tee -a /etc/fstab
Finally, save the RAID configuration so the initramfs can assemble the array early in the boot process:
sudo mdadm --detail --scan | sudo tee -a /etc/mdadm/mdadm.conf
sudo update-initramfs -u # Debian/Ubuntu
# sudo dracut -f # RHEL/CentOS/Fedora
Step 6: Verify Redundancy and Test Failure Recovery
A RAID array is only as good as its ability to survive a disk loss. Simulate a failure by marking one member as faulty and removing it:
# Mark /dev/sdb1 as failed
sudo mdadm /dev/md0 --fail /dev/sdb1
# Remove it from the array
sudo mdadm /dev/md0 --remove /dev/sdb1
The array should transition to a degraded state but remain accessible. Check:
cat /proc/mdstat
Now re‑add the same disk (or a replacement) to confirm rebuilding works:
sudo mdadm /dev/md0 --add /dev/sdb1
Watch the rebuild progress with cat /proc/mdstat. Once it reports clean, the test is successful.
Common Mistakes to Avoid
1 Skipping backups. RAID is not a substitute for regular backups; a double‑disk failure or user error can still erase data.
2 Mixing drive sizes. While you can combine different capacities, the array will only use the smallest drive’s size as the baseline, wasting space.
3 Using the wrong partition type. Forgetting to set the partition type to FD00 (Linux RAID) can confuse mdadm during assembly.
4 Not updating mdadm.conf or initramfs. If the configuration isn’t saved, the array may fail to assemble on boot, leaving the system unbootable.
5 Mounting before the array is fully synced. Accessing the filesystem during the initial sync can degrade performance and, in rare cases, cause data inconsistency.
Tips and Tricks
• Use a dedicated SSD for the RAID superblock. Placing the superblock on a fast SSD reduces latency during boot.
• Enable write‑back caching carefully. For databases, consider hdparm -W0 /dev/sdx to disable caching on individual drives if you experience corruption.
• Monitor health with SMART. Install smartmontools and schedule smartctl -a /dev/sdx checks to catch impending drive failures early.
• Consider RAID 10 for the best of both worlds. It combines mirroring and striping, offering high performance and redundancy at the cost of 50% usable space.
• Document your array. Keep a small text file on the system (e.g., /etc/raid-info.txt) listing level, devices, and creation date for future reference.
Frequently Asked Questions
Can I convert an existing non‑RAID filesystem to RAID without reinstalling?
Yes, but it requires backing up the data, creating the RAID array, and restoring the backup. Direct conversion is risky and not recommended for production systems.
What’s the difference between software RAID and hardware RAID?
Software RAID is managed by the Linux kernel and mdadm, offering flexibility and portability across hardware. Hardware RAID offloads processing to a dedicated controller, which can improve performance but ties the array to that specific card.
Is RAID 5 still safe for critical data?
RAID 5 can survive a single drive failure, but a rebuild after a failure stresses the remaining drives and may encounter a “write‑hole” issue. For mission‑critical workloads, RAID 6 (double parity) or RAID 10 are safer alternatives.
Conclusion
Setting up a Linux software RAID array with mdadm gives you granular control over redundancy, performance, and recovery procedures. By following the six steps above—installing mdadm, preparing disks, creating the array, formatting, mounting, and testing—you’ll have a resilient storage solution that can withstand hardware hiccups while delivering the speed your applications need. Remember to back up regularly, monitor SMART data, and keep your mdadm.conf up to date. With these practices in place, RAID becomes a reliable pillar of any Linux server’s storage strategy.
Photo by Portraits of Modern Dairy on Unsplash
