On this page

What a disk pool is

A disk pool joins several disks into one folder. Instead of films on one disk and series on another, the apps find one folder, /mnt/pool, with everything in it. It is like a bookcase made of several shelves: you find a book on the bookcase, not on a particular shelf.

How Starboard makes one

Starboard uses mergerfs, a small program that shows the disks' folders as one. Each disk stays a normal disk with its own files on it, so:

  • disks can be different sizes, and can be added one at a time;
  • each disk can still be read on its own, in any Linux computer;
  • if a disk fails, only the files on that disk are missing, not the whole pool.

New files go to the disk with the most free space. Balancing moves whole folders between disks to even them out: read Balance the files.

Pools are not backups

A pool keeps one copy of each file. A backup disk protects the pool:

  • Parity (SnapRAID) works out, every night, what is needed to rebuild any one failed disk. One parity disk protects all the others.
  • A copy (rsync) is a second, readable copy of the folders you choose.

Both are in the same computer, so keep a third copy of what matters somewhere else: three copies, on two kinds of storage, one away from home (the 3-2-1 rule). To choose between parity and a copy, read Use a disk as a backup disk.

Other ways to make a pool

  • Unraid and a NAS's own pool (such as Synology): the pool is made by that system, and Starboard shows it.
  • ZFS and Btrfs: file systems that join disks themselves, with their own copies or parity. They need disks of the same size and more memory. Starboard leaves them alone.

Find out more

This page on GitHub, where the help is written.