mdadm: Practical RAID management for Linux

mdadm: Practical RAID management for Linux

mdadm is the go-to command-line tool for managing Linux software RAID arrays. This guide focuses on common tasks, with concrete examples and quick tips for beginners and intermediate users.

Quick start: what you can do

  • Create an array
  • Stop or fail a disk
  • Remove or add disks
  • Inspect array details
  • Clear RAID metadata on a disk

Below are representative commands. Adapt the device paths to your system. Run as root or with sudo where indicated.

Create an array

Create a new RAID array with a specific level and number of disks. Replace the placeholders with your actual devices.

sudo mdadm --create /dev/md/MyRAID --level raid_level --raid-devices number_of_disks /dev/sdXN

Notes:

  • Replace /dev/md/MyRAID with your desired array device (often /dev/md0).
  • Replace raid_level with values like 0, 1, 5, 6, 10, etc.
  • Replace number_of_disks with the number of disks in the array.
  • Replace /dev/sdXN with the actual disk partitions to include (e.g., /dev/sdb1 /dev/sdc1).
  • The system may complain if devices have existing partitions or metadata.

Common pitfall:

  • Don’t mix disks with existing data unless you mean to overwrite. mdadm —create initializes new metadata; data on disks will be lost.

Stop an array

If you need to safely stop an array (for maintenance or replacement):

sudo mdadm --stop /dev/md0

Notes:

  • Stopping does not power down disks; it just stops the kernel from using the array.
  • You may also need to remove the array from the mdadm config after stopping.

Mark a disk as failed

When a disk is suspected to be failing, mark it as failed to prevent syncing on it.

sudo mdadm --fail /dev/md0 /dev/sdXN

Notes:

  • Use with caution; failing a disk will typically trigger a rebuild if redundancy allows.
  • Ensure you have hot-spare disks configured if you rely on automatic rebuilding.

Remove a disk from an array

If a disk is failed or you need to remove it from the array:

sudo mdadm --remove /dev/md0 /dev/sdXN

Notes:

  • The disk must be removed from the array before you can repurpose it.
  • After removal, you might want to zero the superblock on the disk.

Add a disk to an array

To re-add or assemble a disk into an existing array:

sudo mdadm --assemble /dev/md0 /dev/sdXN

Notes:

  • This is commonly used to re-add a hot spare or a rebuilt disk.
  • The disk must have compatible metadata or be prepared for assembly.

Show RAID info

Get a detailed snapshot of an array’s status and configuration:

sudo mdadm --detail /dev/md0

What you’ll see:

  • Array state, raid level, number of devices, and which disks are active/failed.

Reset disk by deleting RAID metadata

If a disk has old or conflicting RAID metadata, you can wipe it:

sudo mdadm --zero-superblock /dev/sdXN

Notes:

  • This does not touch partitions outside the RAID metadata. Use with care if the disk holds other data.

Common pitfalls to avoid

  • Mixing different RAID levels or devices not intended for the same array can corrupt data.
  • Running mdadm commands on the wrong device path (e.g., mistaking /dev/sdX for a partition) can cause data loss.
  • Not updating the kernel initramfs or mdadm config may cause the array not to assemble after reboot. Run update commands as needed for your distro.

Quick maintenance tips

  • Regularly check array health:
    • sudo mdadm --detail /dev/md0 to verify status.
  • Keep a spare drive on hand for hot-replcements if your setup supports it.
  • Keep a backup of critical data; RAID is not a substitute for backups.

Where to learn more

End-to-end example (typical workflow)

  1. Create a 2-disk RAID 1 array:
sudo mdadm --create /dev/md0 --level 1 --raid-devices 2 /dev/sdb1 /dev/sdc1
  1. Check status:
sudo mdadm --detail /dev/md0
  1. Simulate a failure for testing:
sudo mdadm --fail /dev/md0 /dev/sdb1
  1. Remove failed disk:
sudo mdadm --remove /dev/md0 /dev/sdb1

This guide should give you a solid starting point for typical mdadm tasks without diving into advanced features. As you gain confidence, you can explore more options and automations for your specific hardware and redundancy requirements.

See Also