Replacing drives with same drives fails.

[Doralitze]

Hi,
I've got a zpool containing some drives with 512B block size configured. As we've recently swapped out a broken HBA, ZFS now detects that the drives support 4096B block sizes. As such, the pool complains about the incorrect drive sizes:

  pool: massstorage
 state: ONLINE
status: One or more devices are configured to use a non-native block size.
	Expect reduced performance.
action: Replace affected devices with devices that support the
	configured block size, or migrate data to a properly configured
	pool.
  scan: resilvered 2.83G in 00:05:47 with 0 errors on Thu Dec 18 14:38:52 2025
config:

	NAME           STATE     READ WRITE CKSUM
	massstorage    ONLINE       0     0     0
	 raidz1-0     ONLINE       0     0     0
	   sdc_crypt  ONLINE       0     0     0  block size: 512B configured, 4096B native
	   sdd_crypt  ONLINE       0     0     0  block size: 512B configured, 4096B native
	   sdb_crypt  ONLINE       0     0     0  block size: 512B configured, 4096B native
	   sde_crypt  ONLINE       0     0     0  block size: 512B configured, 4096B native

errors: No known data errors

Except for the block size mismatch, the pool works flawlessly. However, replacing (as requested by the above output) the drives, fails with the following error message:

➜  ~ sudo zpool replace massstorage sdc_crypt sdf_crypt
cannot replace sdc_crypt with sdf_crypt: The new device cannot have a higher alignment requirement than the top-level vdev.

The replacement drives model and type are identical to the existing ones. Both drives report an alignment of 0.

Does anyone have an idea on how to fix this issue?

Edit:
I'm using OpenZFS version 2.3 (zfsutils-linux 2.3.2-2).

[amotin]

Try setting ashift pool property and/or zfs_vdev_max_auto_ashift module paremeter to 9. Though since you haven't specified a ZFS version, I suspect it might already be fixed in newer versions.

[Doralitze]

I'm running OpenZFS version 2.3 (zfsutils-linux 2.3.2-2). However running sudo zpool replace -o ashift=9 -f massstorage sdc_crypt sdf_crypt results in the same error response. The same goes for running

# echo "9" > /sys/module/zfs/parameters/zfs_vdev_max_auto_ashift
# zpool replace massstorage sdc_crypt sdf_crypt

[Doralitze]

Which version of ZFS would I need to upgrade to in order to benefit from the fix?

[amotin]

2.3.6 or 2.4.0. Though I am surprised that the module parameter didn't work, so I am not sure.

[Doralitze]

OK, a few weeks later: Both versions 2.3.6 and 2.4.0 didn't do the trick. I ended up creating a new pool and using zfs send / migrate to copy the data.