FCOS BTRFS Raid1 setup

Ask Fedora
1 
┌───────────────────────────────────────────────────────┬─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│                        Design                         │                                                                 Result                                                                  │
├───────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Naive btrfs-native RAID1 root (2× storage.luks +      │ ❌ Fatal. First boot works; every reboot hangs forever — Job dev-disk-by-uuid-…root start running (no limit). FCOS only wires one LUKS  │
│ multi-device btrfs, the current m1pro.bu draft)       │ device into the initramfs (rd.luks.name), so the 2-device btrfs can't assemble at boot. Not fixable without a custom dracut module.     │
├───────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ mdraid1 + single LUKS/TPM2 + btrfs root               │ ✅ Proven. Fresh install, first boot, reboot ×2, btrfs scrub (no errors), ostree redeploy+reboot — all clean. rd.luks.name=…=root       │
│ (boot_device.mirror sugar)                            │ single entry, clevis TPM2 unlock every boot, md126/md127 both [UU], full-size 11.5 G root (no partition-shrink bug). ESP + /boot        │
│                                                       │ mirrored too.                                                                                                                           │
└───────────────────────────────────────────────────────┴─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┘

i know deafult is diffrent, but somehow i prefer BTRFS more as it gives more functions and features what i need and makes setup more better. Shame BTRFS native RAID1 fails terrible so had to go difrent way

2

I don’t use coreos, but it is my understanding that the combination of btrfs and raid has problems that seem to discourage using raid with btrfs.

It seems consistent that users claim btrfs is reliable enough in checking integrity and the use of CoW tech to prevent failure in a catastrophic way. A compromised file system seems to mount read only (ro) so there is usually no loss of data and recovery seems usually to be relatively simple.