Btrfs RAID 5/6: Why You Still Shouldn't Use It in Production in 2026
Btrfs as a filesystem with checksums and snapshots is ideal for storage. But Btrfs RAID 5/6 has had the "experimental" label for years – and for good reason. Here's the honest 2026 status.
Short version
Btrfs RAID 5/6 has the infamous write hole problem: on a power loss mid-write, parity can become inconsistent and corrupt data is silently restored. Plus: scrub repairs can damage data themselves. Even Btrfs maintainers have said for years "not for production". Recommendation: avoid Btrfs RAID 5/6. Use ZFS RAIDZ2 or classic mdadm RAID 5/6 with Btrfs as the filesystem on top.
What is the write hole?
In RAID 5/6 a file plus its parity must be written together. If power dies mid-write, half (data or parity) might land on disk, the other not. On read, the system doesn't see the inconsistency – the corrupted data is treated as "correct".
ZFS solves this with copy-on-write and transactional writes. mdadm RAID 5/6 partially solves it via journal devices. Btrfs RAID 5/6 has the problem structurally open.
Known bugs 2024-2026
- Scrub can degrade data: on a scrub with damaged sectors the correct copy sometimes gets overwritten. Documented since 2018, not fully solved.
- Balance can hang: Btrfs balance on RAID 5/6 with large data can hang for days or freeze the system.
- Subvolume deletion slows exponentially with larger RAID 5/6 pools.
- Read errors can't be reliably recovered when both data copies differ from parity.
Source: official Btrfs wiki status page, kernel.org mailing list 2024-2025.
What Btrfs still does well
Important: the issue is only Btrfs RAID 5/6. Btrfs in other modes is mature and good:
- Btrfs RAID 1 / RAID 10: stable, production-ready, Synology has been using it for years.
- Btrfs single (on mdadm RAID 5): mdadm provides RAID, Btrfs adds snapshots, checksums, compression. Best practice for many setups.
- Btrfs snapshots: excellent, fast, low overhead.
- Btrfs compression (zstd): saves 20-50% on text/logs without noticeable CPU cost.
Stable alternatives for RAID 5/6
ZFS RAIDZ2. Gold standard for 4+ drives. No write hole, block checksums, auto-repair. More in our Bitrot & ZFS guide.
mdadm RAID 5/6 + Btrfs single. Linux classic. mdadm does RAID, Btrfs (or ext4) does the filesystem. Stable, widely used.
Synology SHR-1/SHR-2. Uses mdadm + Btrfs underneath. Synology's packaging makes it stable out of the box.
Unraid. Own approach with per-drive filesystem and a dedicated parity drive. Not classic RAID 5/6, its own model.
When (if ever) Btrfs RAID 5/6 makes sense
- Test/lab where data loss is acceptable
- Setups with UPS plus daily off-site backup as safety net
- If you follow the Btrfs mailing list and are willing to report bugs
For home NAS or production: no.
Migrating away from Btrfs RAID 5/6
If you're running Btrfs RAID 5/6 today:
- Full backup (two copies) to a different medium
- Destroy the pool
- Create a new pool as ZFS RAIDZ2 or mdadm + Btrfs
- Restore data from backup
Duration: 1-7 days depending on data volume. Annoying but necessary for integrity.
Recommendation
If you want Btrfs: stick to RAID 1/10 or Btrfs single on mdadm. If you want RAID 5/6: ZFS RAIDZ2 (with bitrot protection) or mdadm RAID 6 (classic). Btrfs RAID 5/6 in 2026 remains "not for production".
Related articles
Further reading
RAID 0 Explained: Speed Without a Safety Net
RAID 5 vs RAID 6: Which Should You Actually Pick?