RAID 9 is not a standard RAID level supported by Linux mdadm, Windows Disk Management, or macOS. It likely refers to RAID-10 (mirrored stripes) or a proprietary vendor-specific configuration. Using three disks for RAID-10 is impossible as it requires an even number of drives. For three 500GB disks, RAID-5 or ZFS RAID-Z provides optimal redundancy and capacity.
Understanding the RAID 9 Misconception
The term "RAID 9" often appears in user queries due to confusion with other RAID levels or specific hardware controller labels. Standard RAID levels defined by the industry include RAID 0 (striping), RAID 1 (mirroring), RAID 5 (distributed parity), and RAID 6 (double distributed parity). There is no universally accepted "RAID 9" in software-based solutions like mdadm on Linux or built-in OS utilities. Some older hardware controllers might label proprietary configurations differently, but this is rare and non-standard.
In many contexts, users mistakenly refer to RAID-10 (also known as RAID 1+0) when they say "RAID 9." RAID-10 combines mirroring and striping to provide high performance and redundancy. However, RAID-10 requires an even number of disks because it creates pairs of mirrored disks and then stripes across those pairs. With only three disks, you cannot create a true RAID-10 array. Attempting to force a four-disk configuration with three available drives will fail.
Why Three Disks Cannot Do RAID-10
RAID-10 works by creating multiple RAID-1 mirrors and then striping data across them. Each mirror pair requires two disks. Therefore, you need at least two disks for one mirror, four disks for two mirrors, six disks for three mirrors, and so on. With three disks, you have an odd number, which breaks the fundamental requirement of RAID-10. You would end up with one disk left over, which cannot be utilized in a standard RAID-10 setup without leaving it unused or using a complex and unsupported workaround.
Recommended Alternatives for Three Disks
Given the constraint of three 500GB disks, the most robust and widely supported solution is RAID-5 or ZFS RAID-Z. Both offer similar benefits: they use one disk's worth of space for parity, allowing you to lose one disk without data loss, while providing increased read performance through striping.
Option 1: Linux mdadm RAID-5
If you are using Linux, mdadm is the standard tool for managing software RAID arrays. Here is how to set up a RAID-5 array with three 500GB disks:
- Identify your disks: Use
lsblkordmesg | grep sdto find the device names (e.g., /dev/sdb, /dev/sdc, /dev/sdd). - Create partitions: Ensure all three disks have identical partition tables. You can use
fdiskorpartedto create a single primary partition on each disk. - Create the array: Run the following command:
sudo mdadm --create /dev/md0 --level=5 --raid-devices=3 /dev/sdb1 /dev/sdc1 /dev/sdd1. - Format the array: Once created, format it with a filesystem like ext4:
sudo mkfs.ext4 /dev/md0. - Mount the array: Create a mount point and mount the array:
sudo mkdir /mnt/raid5 && sudo mount /dev/md0 /mnt/raid5. - Save configuration: Save the array configuration to ensure it loads on boot:
sudo mdadm --detail --scan | sudo tee -a /etc/mdadm/mdadm.conf.
This setup gives you approximately 1TB of usable storage (500GB * 2) with protection against one disk failure.
Option 2: ZFS RAID-Z
ZFS is a modern filesystem that includes RAID capabilities. It is highly recommended for its data integrity features and ease of management. To create a RAID-Z array with three disks:
- Install ZFS: On Ubuntu/Debian, install
zfsutils-linux. On CentOS/RHEL, installzfs. - Create the pool: Use the command
sudo zpool create mypool raidz /dev/sdb /dev/sdc /dev/sdd. - Verify: Check the status with
zpool statusandzpool list.
ZFS automatically handles parity and offers snapshotting and compression out of the box. The usable space will also be around 1TB.
Performance and Redundancy Considerations
Both RAID-5 and ZFS RAID-Z provide excellent redundancy for small setups. They allow you to continue operating if one disk fails, though performance may degrade during rebuilds. It is crucial to replace failed disks promptly to restore redundancy.
Read performance is generally improved due to striping, where data is read from multiple disks simultaneously. Write performance may be slightly slower than RAID-0 due to parity calculations, but it remains efficient for most workloads.
If you need higher performance and can add more disks, consider expanding to four disks for RAID-10 or RAID-Z2 (which allows two disk failures). However, with three disks, RAID-5/RAID-Z is the best balance of capacity, redundancy, and simplicity.
For more information on configuring RAID with external drives, see our guide on Is it possible to make a RAID of multiple WD Elements external drives?. Additionally, if you are interested in optimizing RAID-1 for maximum bandwidth, check out How to Optimize RAID-1 for Maximum Bandwidth in Sequential Write/Read Processing.
Frequently Asked Questions
Can I create a RAID-10 array with three disks?
No, RAID-10 requires an even number of disks because it relies on mirrored pairs. With three disks, you cannot form complete mirror pairs, making RAID-10 impossible to implement.
What happens if one disk fails in a RAID-5 array?
The array continues to operate in a degraded mode. You can still read and write data, but performance may drop. You must replace the failed disk and rebuild the array as soon as possible to restore full redundancy.
Is ZFS better than mdadm for three disks?
ZFS is often preferred for its advanced data integrity features, such as checksumming and automatic repair, whereas mdadm is simpler but lacks some of these protections. For most users, ZFS provides a more robust solution.
How much usable space do I get with three 500GB disks in RAID-5?
You will get approximately 1TB of usable space. One disk's worth of capacity (500GB) is used for parity, leaving two disks' worth for data storage.
Can I upgrade from RAID-5 to RAID-6 later?
With ZFS, you can easily expand your pool by adding more disks and converting to RAID-Z2. With mdadm, upgrading from RAID-5 to RAID-6 is complex and may require backing up data, recreating the array, and restoring data.