Software RAID: Setup, Management, and Administration

Below is a full, detailed, production-ready guide for Software RAID Setup, Management, Monitoring, and Drive-Failure Recovery.

What is Software RAID?

Software RAID is a method of combining multiple physical drives into a single logical unit using the operating system (not a dedicated hardware controller). It’s commonly implemented with tools like mdadm on Linux.

What RAID Actually Does (Core Idea)

RAID = Redundant Array of Independent Disks

At its core, RAID uses 3 fundamental mechanisms:

1. Striping (Performance)

   Data is split into chunks and written across multiple disks.

   • Improves read/write speed
   • No redundancy (if one disk fails → all data lost)
   • Used in RAID 0

2. Mirroring (Redundancy)

   Data is duplicated across disks.

   • Provides fault tolerance
   • If one disk fails, data still exists on another
   • Used in RAID 1

3. Parity (Recovery)

   Extra data (parity) is calculated so lost data can be rebuilt.

   • Enables data reconstruction
   • More storage-efficient than mirroring
   • Used in RAID 5, RAID 6

⚙️ How Software RAID Works (Under the Hood)

Unlike hardware RAID, everything is handled by the OS kernel:

• ✔️ Logical Layer (Virtual Device)

  • OS creates a virtual device like:

    /dev/md0
  

  • This acts like a normal disk (you can format it, mount it, etc.)

• ✔️ RAID Engine (Kernel + mdadm)

  • The Linux kernel RAID subsystem (md) handles:
  • Splitting data into stripes
  • Writing mirrors
  • Calculating parity (XOR operations)
  • mdadm is just the management tool (create, assemble, monitor)

• ✔️ Block-Level Operations

  When an app writes data:

  1. OS receives write request
  2. RAID layer intercepts it
  3. Data is:
  • Split (striping)
  • Duplicated (mirroring)
  • Or parity-calculated
  1. Written to multiple disks accordingly

• ✔️ Example (RAID 5 write)

  Write request: DATA = A + B + C

  • Disk 1 → A
  • Disk 2 → B
  • Disk 3 → C
  • Disk 4 → Parity (A ⊕ B ⊕ C)

  If Disk 2 fails → B can be rebuilt using:

  B = A ⊕ C ⊕ Parity
  

Key RAID Levels in Software RAID

Software RAID vs Hardware RAID

Why Use Software RAID (Common in VPS / Linux Servers)

• No need for RAID cards
• Works great with modern CPUs
• Fully scriptable and automatable
• Easier recovery in many cases

🧩 Summary

Software RAID is:

• OS-driven disk aggregation
• Built on striping, mirroring, and parity
• Managed via tools like mdadm
• Extremely common in Linux hosting environments

This applies to most Linux server environments including cPanel/WHM, AlmaLinux/Rocky, Ubuntu/Debian, and generic VPS/dedicated servers—and uses mdadm, the standard Linux software RAID manager.

Software RAID Setup, Management & Administration

(with Full Failure-Recovery Procedures)

1. Getting Started with Software RAID

   Software RAID uses the OS kernel (via mdadm) to create and manage arrays of multiple disks for redundancy, performance, or both.

   Common RAID Levels

   Level	Min Disks	Purpose	Fault Tolerance	Notes
   RAID 0	2	Stripe	0	Performance only; not recommended for production
   RAID 1	2	Mirror	1 disk	Most common for servers (OS partitions)
   RAID 5	3	Stripe + parity	1 disk	Good compromise; slow rebuild; not recommended on large disks
   RAID 6	4	Stripe + dual parity	2 disks	Better for large disks
   RAID 10	4	Striped Mirrors	1+	Best performance + redundancy

   mdadm software RAID is extremely common on WHM/cPanel dedicated servers and Linux VPS.

2. Install Required Tools

   • RHEL / AlmaLinux / Rocky

     sudo dnf install mdadm smartmontools -y
     

   • Debian / Ubuntu

     sudo apt install mdadm smartmontools -y
     

3. Create a New RAID Array

   1. Identify the Disks

      lsblk
      fdisk -l
      

      Assume disks /dev/sdb and /dev/sdc.

   2. Prepare Disks (create partitions)

      Use GPT for modern layouts:

      parted /dev/sdb mklabel gpt
      parted /dev/sdb mkpart primary 0% 100%
      parted /dev/sdc mklabel gpt
      parted /dev/sdc mkpart primary 0% 100%
      

      Mark partitions as RAID:

      sudo parted /dev/sdb set 1 raid on
      sudo parted /dev/sdc set 1 raid on
      

      Partitions become:

      /dev/sdb1
      /dev/sdc1
      

4. Create RAID Arrays

   1. RAID 1 (Mirror)

      mdadm --create /dev/md0 --level=1 --raid-devices=2 /dev/sdb1 /dev/sdc1
      

   2. RAID 5 Example

      mdadm --create /dev/md0 --level=5 --raid-devices=3 /dev/sd[bcd]1
      

5. Add Filesystem & Mount

   mkfs.ext4 /dev/md0
   mkdir /mnt/raid
   mount /dev/md0 /mnt/raid
   

6. Persist RAID Assembly

   Write array definition:

   mdadm --detail --scan >> /etc/mdadm.conf
   

   Or on Debian-based:

   mdadm --detail --scan >> /etc/mdadm/mdadm.conf
   update-initramfs -u
   

7. Monitoring RAID Health

   1. Check RAID Status

      cat /proc/mdstat
      

      Sample output:

      md0 : active raid1 sdb1[0] sdc1[1]
            976630336 blocks [2/2] [UU]
      

      • UU = both disks healthy
      • _U = left disk failed
      • U_ = right disk failed

   2. Detailed View

      mdadm --detail /dev/md0
      

8. Replacing a Failed Drive (RAID1/5/6/10)

   This is the most important part for production systems.

   Symptoms of Failed Disk

   • cat /proc/mdstat shows _U or U_
   • Server logs show I/O errors
   • SMART failures: smartctl -a /dev/sdX

9. Step-by-Step Drive Failure Recovery

   Assume:

   • Array: /dev/md0
   • Bad disk: /dev/sdb1
   • Replacement disk: /dev/sdd

   1. Identify the Faulty Drive

      mdadm --detail /dev/md0
      

      You’ll see something like:

      Number  Major  Minor  RaidDevice State
         0     8       17        0      faulty   /dev/sdb1
         1     8       33        1      active   /dev/sdc1
      

   2. Mark Drive as Failed

      mdadm --fail /dev/md0 /dev/sdb1
      

   3. Remove the Failed Drive

      mdadm --remove /dev/md0 /dev/sdb1
      

   4. Prepare the New Drive

      If whole disk:

      parted /dev/sdd mklabel gpt
      parted /dev/sdd mkpart primary 0% 100%
      parted /dev/sdd set 1 raid on
      

   5. Add New Drive to the Array

      mdadm --add /dev/md0 /dev/sdd1
      

      Rebuild begins automatically.

      Monitor progress:

      watch cat /proc/mdstat
      

10. Rebuilding the Array

    Expected output during rebuild:

    [>....................]  rebuild = 5.3% (103424/1953512448) finish=120.5min speed=26000K/sec
    

11. Clone Partition Table Automatically (Optional Best Practice)

    If your drives must match exactly:

    sfdisk -d /dev/sdc | sfdisk /dev/sdd
    

    Then add the partition:

    mdadm --add /dev/md0 /dev/sdd1
    

12. Hot Spare Setup (Automatic Recovery)

    Add a spare disk:

    mdadm --add /dev/md0 /dev/sde1
    

    Verify:

    Spare Devices : 1
    

    If a disk fails, mdadm automatically pulls in the spare.

13. SMART Monitoring

    Schedule SMART tests:

    Create /etc/cron.weekly/smartcheck:

    #!/bin/bash
    smartctl -t short /dev/sda
    smartctl -t short /dev/sdb
    

14. 📨Email Alerts for RAID Failure

    Install mdadm mail alerts:

    Edit /etc/mdadm.conf:

    MAILADDR admin@yourdomain.com
    

    Restart:

    systemctl restart mdmonitor
    

15. Advanced Diagnostics

    • Check current RAID bitmap (helps fast rebuild)

      mdadm --detail /dev/md0 | grep -i bitmap
      

    • Verify stripes (RAID5/6)

      echo check > /sys/block/md0/md/sync_action
      

16. Troubleshooting Scenarios

    1. Scenario A: RAID shows “degraded” even after rebuild

       Force re-add disk:

       mdadm --add /dev/md0 /dev/sdd1 --force
       

    2. Scenario B: md0 will not assemble on boot

       mdadm --assemble --scan
       

    3. Scenario C: Accidentally removed the wrong disk

       Re-add it:

       mdadm --add /dev/md0 /dev/sdb1
       

    4. Scenario D: Superblock errors

       Zero superblock before reuse:

       mdadm --zero-superblock /dev/sdd1
       

17. Backup mdadm metadata (critical!)

    Save RAID definition:

    mdadm --detail --scan > /root/mdadm.backup
    

    Save disk partition tables:

    sfdisk -d /dev/sda > /root/sda.part
    sfdisk -d /dev/sdb > /root/sdb.part
    

18. Full Cleanup Commands (Destroy RAID)

    umount /mnt/raid
    mdadm --stop /dev/md0
    mdadm --remove /dev/md0
    mdadm --zero-superblock /dev/sd[bcd]1
    

Summary: Best Practices for Software RAID Administration

✔ Always use RAID 1 or RAID 10 for critical servers
✔ Keep at least one hot spare on RAID 5/6/10
✔ Enable email alerts
✔ Monitor smartctl logs weekly
✔ Run periodic RAID checks
✔ Save /etc/mdadm.conf after any modification
✔ Use identical disks whenever possible
✔ Keep a replacement drive on hand

Conclusion

You now know about software RAID setup, management, and administration.