ZFS, Actually Storage you can finally reason about.

RAIDZ Expansion

definition
depth 2/2 − deeper + surface

Can you add one more disk to an existing RAIDZ team when you need more room? Two confident answers circulate: "of course—storage grows" (what store-bought boxes teach) and "never—ZFS locks the team size forever" (a decade of forum truth). Actually, both are stale: since early 2025 (OpenZFS version 2.3), you can "add new devices to an existing RAIDZ pool, increasing storage capacity without downtime" (release notes)—one command, while everything keeps running.

When reading ZFS advice, check the date. This feature made years of confident guidance obsolete overnight, and the forums have long memories.

The caveats that keep it from being magic:

  • Existing data keeps its old parity-to-data ratio. Expansion reflows blocks across the new width, but stripes written as 3-wide keep 3-wide economics until naturally rewritten—so reclaimed capacity is less than the naive math promises, converging only as data churns. (Carried at medium confidence from the feature's design discussion—vet against a primary source before it bears operational weight.)
  • Type is still fixed. RAIDZ1 cannot become RAIDZ2 in place; parity level remains a creation-time decision.
  • No shrink. Width goes up only.
  • The expansion itself is a long, resilver-like background reflow.

The planning consequence: expansion is a real escape hatch for "I need one more disk of room," not a substitute for choosing vdev shape deliberately at creation—the pressure the old absolute created was mostly good pressure.