Skip to content

The Neighbours

MergerFS is one answer among many to "my files are spread across drives." These are the neighbours, and what each trades. Nothing here is a ranking; the right tool is the one whose failure mode you can live with.

Tool Shape Choose it when
MergerFS Userspace (FUSE) union of plain folders; files whole on one drive Your drives differ, arrive one at a time, and must stay readable on their own
RAID 0 Kernel striping across drives You want one fast volume and accept that one dead drive takes everything
RAID 1 / 5 / 6 (mdadm) Mirroring or parity in the kernel You want redundancy and one volume, with matched drives and rebuilds you understand
LVM Volume manager: pools physical drives into logical volumes, striping or linear You want resizing and snapshots more than per-file placement, on drives you'll manage as a set
ZFS Filesystem and volume manager in one; pools (vdevs) with checksums and parity You are building the storage properly from the start and can give it the drives (and RAM) it asks for
SnapRAID Scheduled parity over standalone drives Your data changes slowly and you want drive-level protection; the classic partner under mergerfs, not a rival
UnionFS / OverlayFS Kernel union mounts, built for container image layers You are stacking a writable layer over read-only ones — images, live systems — not pooling data drives
Windows: Storage Spaces / DrivePool The platform equivalents Your drives live on a Windows box; DrivePool in particular is mergerfs's closest cousin
Just mount the drives and keep a list No software at all Honestly, for two drives and a tidy mind — until the third drive arrives

MergerFS's own trade is visible in the table: it is the only entry that leaves every drive independently readable and demands nothing of the drives — no matching sizes, no formatting, no commitment — and the only cost it refuses to hide is that it protects nothing. RAID and ZFS make the opposite trade: real redundancy, real lock-in, drives that mean nothing alone.

Two honest notes. Moving between these is mercifully easy in one direction: leaving mergerfs means unmounting it and owning some folders — an afternoon, not a migration. Moving to it from a striped array is the usual backup-restore project. And the question every row of the table must answer is chapter 10's: when a drive dies, what exactly do you lose, and what exactly brings it back?