How to Build a Software RAID1 on CentOS 7 Using mdadm – Step‑by‑Step Guide

RAID (Redundant Arrays of Independent Disks) provides high availability and redundancy, but hardware RAID cards can be costly for large deployments; software RAID offers a cheaper alternative.

Implementation Steps

The experiment uses a CentOS 7.2 VM with two 50 GB data disks (host1). In production, RAID disks should be identical in brand, model, and capacity.

1. Verify mdadm installation

rpm -qa | grep mdadm

Output shows mdadm-3.3.2-7.el7.x86_64.

2. Partition the disks and set type to RAID

Use fdisk for disks ≤2 TB (type code fd) or parted for larger disks, creating a GPT label and setting the partition flag to RAID.

3. Create RAID1 with mdadm

Run:

mdadm -C /dev/md0 -ayes -l1 -n2 /dev/xvdb1 /dev/xvdc1

The note indicates software RAID cannot be used as a boot device.

Command options:

-C --create: create array

-a --auto: auto‑create device (use -a yes)

-l --level: array level (e.g., raid1)

-n --raid-devices: number of active disks

/dev/md0: array device name

Check status with:

cat /proc/mdstat

Initial resync may show ~95 % complete; a second check confirms full synchronization.

Alternatively, view details with:

mdadm -D /dev/md0

4. Create mdadm configuration file

echo DEVICE /dev/sd{a,b}1 >> /etc/mdadm.conf mdadm -Evs >> /etc/mdadm.conf

mdadm reads this file at startup to assemble arrays automatically.

5. Use the RAID device

Format and mount:

mkfs.ext4 /dev/md0

Optionally specify stride for performance, e.g.:

mkfs.ext4 -E stride=16 -b 4096 /dev/md0

6. Additional commands

Manage array health:

Mark a failed disk: mdadm /dev/md1 -f /dev/sdb5 Remove a failed disk: mdadm /dev/md1 -r /dev/sdb5 Add a new disk: mdadm /dev/md1 -a /dev/sdb7 Stop the array:

mdadm -S /dev/md1