ZFS AnyRaid: Mixed-Size Drives Finally Come to ZFS

If you’ve ever tried to build a ZFS NAS on a budget, you’ve probably run into the mixed-drive problem. You’ve got a 4TB drive from an old PC, a 6TB you grabbed on sale, and maybe a fresh 8TB for your growing media collection. In the ZFS world, those extra terabytes on your larger drives? Dead space. Wasted.

For years, the advice has been simple: buy matching drives or accept the loss. But that advice ignores reality. Real people upgrade incrementally. Real people shop sales. Real people repurpose old hardware.

ZFS AnyRaid is here to fix that.

The Problem: ZFS Hates Your Mismatched Drives

Traditional ZFS has a simple but brutal rule: all drives in a VDEV operate at the size of the smallest disk. This isn’t a bug—it’s a fundamental design choice that dates back to ZFS’s enterprise origins.

Let’s say you’ve got this setup:

Pool: mediavault
├── vdev: raidz1
│   ├── sda: 3TB
│   ├── sdb: 3TB  
│   ├── sdc: 5TB  ← 2TB wasted
│   └── sdd: 5TB  ← 2TB wasted

In a traditional RAID-Z1, ZFS treats this as four 3TB drives. Those 5TB drives might as well be 3TB units. You’ve paid for 16TB of raw storage, but you’re only using 12TB.

# What you'd see with traditional ZFS
$ zpool list -v mediavault
NAME        SIZE   ALLOC   FREE
mediavault  10.8T  4.2T    6.6T
  raidz1    10.8T  4.2T    6.6T

Want to upgrade? You can’t just swap one drive. You need to replace all drives in the VDEV before you see any capacity increase. That’s the kind of upgrade that requires planning, budget, and patience—luxuries many homelabbers don’t have.

Why does ZFS do this? It’s about predictability. ZFS stripes data across all drives in a VDEV, and that stripe pattern is set at VDEV creation. Changing it requires rewriting everything. Enterprise environments plan storage in matched sets, so this wasn’t a priority—until now.

Enter AnyRaid: Tiles Change Everything

ZFS AnyRaid introduces a completely new way to think about storage allocation. Instead of treating each physical disk as a monolithic unit, AnyRaid breaks drives into 64GB “tiles”.

Here’s how it works:

The Tile Concept

Physical Drives                    Logical Tiles
┌─────────────────┐               ┌──┬──┬──┬──┬──┬──┐
│   6TB Drive     │    ───────►   │T1│T2│T3│...│95│96│  (96 tiles)
└─────────────────┘               └──┴──┴──┴──┴──┴──┘

┌─────────────────┐               ┌──┬──┬──┬──┬──┐
│   3TB Drive     │    ───────►   │T1│T2│T3│...│48│    (48 tiles)
└─────────────────┘               └──┴──┴──┴──┴──┘

Each tile becomes a logical unit that ZFS can manage independently. When you write data, ZFS allocates tiles across your drives, ensuring redundancy by mirroring tiles on different physical devices.

Why 64GB?

The tile size is a calculated trade-off:

• Small enough to allow fine-grained allocation across mixed drives
• Large enough to keep metadata overhead manageable
• Divisible efficiently across common drive sizes

# Conceptual view of tile allocation
# Configuration: 1x6TB + 2x3TB drives with AnyRaid-Mirror
#
# Total tiles: 96 + 48 + 48 = 192 tiles
# With 2-way mirror: 96 mirrored pairs
# Usable capacity: ~6TB (with full redundancy)
#
# Traditional ZFS mirror with same drives: ~3TB usable
# AnyRaid gain: +100% capacity!

The Indirection Layer

The magic happens through a new indirection map—an on-disk structure that tracks which logical tiles map to which physical locations. This extra layer enables:

1. Flexible placement — Tiles can go anywhere there’s space
2. Automatic rebalancing — New drives get utilized immediately
3. Preserved redundancy — ZFS always knows where your data’s copies live

Important: The indirection layer does add some computational overhead. For most homelab workloads (media streaming, file serving, backups), this overhead is negligible. Heavy write-intensive workloads may see minor performance differences compared to traditional layouts.

Capacity Gains: The Numbers Game

Let’s look at a realistic homelab scenario.

Scenario: The Budget Builder

You’re building a NAS with drives collected over time:

Drive	Capacity	Source
1	4TB	Old PC backup
2	6TB	Black Friday deal
3	6TB	Same deal
4	8TB	Birthday splurge

Total raw capacity: 24TB

Here’s what you get:

Layout	Usable Space	Efficiency
Traditional Mirror (2x2)	10TB	42%
Traditional RAID-Z1	12TB (capped at 4TB/drive)	50%
AnyRaid-Mirror	12TB	50%
AnyRaid-Z1	~16TB	67%

With AnyRaid-Z1, you gain 4TB of usable space compared to traditional RAID-Z1—and you can add more drives incrementally without replacing everything.

Real Financial Impact

Think about what that means in dollars:

# Traditional ZFS upgrade path
# Start: 4x mixed drives (4TB, 6TB, 6TB, 8TB)
# Want: More capacity
# Option: Replace ALL drives with 8TB units
# Cost: ~$320-480 (4x 8TB drives)

# AnyRaid upgrade path
# Start: Same mixed drives
# Want: More capacity
# Option: Add another drive whenever you want
# Cost: ~$80-120 per incremental upgrade

The flexibility to add storage when it makes sense for your budget—not when the VDEV replacement math works out—is a game-changer for home server builders.

Timeline: When Can You Use This?

As of February 2026, AnyRaid is still in active development. There’s no confirmed release date, but the pace has surprised observers.

Development Phases

Phase 1 (Current Focus):

• AnyRaid-Mirror support
• Indirection Map implementation
• VDEV Expansion capability
• Layout Optimization algorithms

Phase 2 (Future):

• AnyRaid-Z1, Z2, Z3 support
• Extended parity layouts

Who’s Building This?

The Team:

• HexOS — Modern home-server OS built on TrueNAS, sponsor of the project
• Klara Systems — ZFS internals experts, lead development
• Allan Jude — Project lead
• Paul Dagnelie — Principal ZFS developer, AnyRaid architect

Jon Panozzo, CEO of HexOS, put it plainly:

“People don’t buy perfectly matched drive sets. They upgrade over time, and ZFS historically punished them for that.”

The development speed has been notable. In an industry where major ZFS features often take 3-5 years from proposal to release, AnyRaid has progressed rapidly since its May 2025 announcement.

# Historical context for major ZFS features
# DRAID: Announced 2015, merged 2020 (~5 years)
# RAID-Z Expansion: Proposed 2017, merged 2022 (~5 years)
# AnyRaid: Announced May 2025, active development (~1 year)

AnyRaid vs Unraid: The Homelab Showdown

Let’s address the elephant in the room. Unraid has owned the mixed-drive homelab market for years. How does AnyRaid change things?

What Unraid Does Well Today

• Native mixed-drive support (always had it)
• Simple drive-by-drive expansion
• Individual drive spin-down for power savings
• Very beginner-friendly UI

What AnyRaid Brings

Feature	Unraid	ZFS + AnyRaid
Mixed drives	✅	✅
Data integrity (checksums)	❌	✅
Native snapshots	❌	✅
Replication	Limited	✅
Compression	❌	✅ (LZ4/ZSTD)
Deduplication	❌	✅
Enterprise support	❌	✅

The real win: AnyRaid gives you Unraid-style flexibility with ZFS’s legendary data integrity. Those checksums that ZFS computes for every block? They’ve saved countless home servers from silent data corruption. Now you can have that safety net and mixed-drive support.

Should You Switch?

Consider AnyRaid if you:

• Want ZFS’s data integrity guarantees
• Value snapshots for backup/rollback
• Have existing ZFS experience
• Plan to use TrueNAS or HexOS

Stick with Unraid if you:

• Need individual drive spin-down (power is a concern)
• Prefer GUI-over-command-line management
• Want something that exists today
• Don’t need ZFS-specific features

The beauty of AnyRaid is that it expands your options—ZFS no longer requires matched drives to make sense for homelab budgets.

Practical Example: Building with AnyRaid

Once AnyRaid lands, here’s how pool management might look:

# Create an AnyRaid-Mirror pool with mixed drives
$ zpool create mediapool anyraid-mirror sda sdb sdc sdd

# Check your capacity utilization
$ zpool list -v mediapool
NAME              SIZE   ALLOC   FREE   DEDUP  HEALTH
mediapool        12.0T  2.4T    9.6T   1.00x  ONLINE
  anyraid-mirror 12.0T  2.4T    9.6T         ONLINE
    sda           4.0T   -       -            ONLINE
    sdb           6.0T   -       -            ONLINE
    sdc           6.0T   -       -            ONLINE
    sdd           8.0T   -       -            ONLINE

# Add a new drive later (instant capacity gain)
$ zpool add mediapool anyraid-mirror sde
$ zpool list mediapool
NAME        SIZE   ALLOC   FREE
mediapool  16.0T  2.4T    13.6T

The key difference from traditional ZFS: you added a drive and immediately got more usable space. No replacement cycle. No wasted capacity period.

What About Reliability?

This is the question every ZFS user asks. Does the indirection layer compromise ZFS’s core promise?

Short answer: No.

Paul Dagnelie, the lead architect, addressed this directly:

“This allows us to preserve the same reliability guarantees that mirrors and RAID-Z provide. If you lose a device, you can still reconstruct all of your data.”

The indirection map is itself mirrored and checksummed. ZFS still maintains multiple copies of your data across different physical drives. Standard resilver processes work. If a drive fails, ZFS knows exactly which tiles need to be rebuilt and where to find their redundant copies.

Caveat: AnyRaid-Mirror provides 2-way redundancy (survive 1 drive failure per mirror group). For critical data, consider AnyRaid-Z2 (survive 2 drive failures) when it becomes available. Always follow the backup rule: important data should exist in multiple places, not just multiple drives in one array.

The Bottom Line for Homelabbers

ZFS AnyRaid represents a fundamental shift in who ZFS is for.

For decades, ZFS has been the enterprise filesystem that power users brought home. It offered unmatched data integrity, but required enterprise-style planning: matched drives, full-VDEV upgrades, careful expansion strategies.

AnyRaid changes that equation:

Before AnyRaid:

• ZFS = Enterprise features + enterprise discipline
• Mixed drives = wasted money and space
• Budget builds = choose Unraid or suffer inefficiency

After AnyRaid:

• ZFS = Enterprise features + consumer flexibility
• Mixed drives = fully utilized capacity
• Budget builds = ZFS is finally viable

The feature isn’t ready yet, but for homelabbers watching from the sidelines, ZFS AnyRaid is worth the wait. It’s the missing piece that makes ZFS accessible to anyone who builds their storage incrementally—which, let’s be honest, is most of us.

---

Have thoughts on AnyRaid or questions about ZFS storage planning? Drop a comment below or reach out on Mastodon.