Hi all,

I switched an existing USB-booted server to internal boot and ended up with a working boot pool, plus a format prompt I'd like to understand before touching anything.

System

- Unraid 7.3.2 (upgraded from 7.2.5 the same day), UEFI boot, Pro licence (still on the USB stick, which stays plugged in)

- ASUS Z690 board, i9-12900K

- Boot devices: 2x Transcend SSD370 32 GB on the onboard SATA ports (ahci)

- Array: 2 parity + 24 data on 2x LSI 9300-16i; two NVMe pools (btrfs)

What I did

Onboarding Wizard -> Internal boot -> Mirrored -> both SSD370 -> boot size 0 (dedicated) -> update BIOS boot order enabled, no reboot.

What happened

The wizard ended with "Internal boot setup timed out, but was probably set up on your server. Please reboot to finish" and:

Unable to resolve boot device for serial 'TS32GSSD370_XXXXXXXX' from assignableDisks; skipping BIOS entry creation for this disk. (same for the second SSD)

Running: emcmd cmdCreatePool=apply&poolName=boot&poolSlots=2

Running: emcmd cmdAssignDisk=apply&diskName=boot&diskId=TS32GSSD370_XXXXXXXX

Running: emcmd cmdAssignDisk=apply&diskName=boot2&diskId=TS32GSSD370_XXXXXXXX

Running: emcmd cmdMakeBootable=apply&poolName=boot&poolBootSize=0

mkbootpool: command failed or timed out

emcmd request failed with status 500

Syslog shows mkbootable itself completed on both SSDs (GPT written, zpool "flash" created, grub installed for x86_64-efi and i386-pc, "No error reported"), followed by:

emhttpd: online: TS32GSSD370_XXXXXXXX (sdab) ... => data partition: sdab4 type: gpt start: 62529536 size: 1860 ... bootable

emhttpd: online: TS32GSSD370_XXXXXXXX (sdac) ... => data partition: sdac4 type: gpt start: 62529536 size: 1860 ... bootable

emhttpd: debug: cmdMakeBootable: not boot eligible

emhttpd: HTTP/1.1 500 Internal Server Error

Despite "skipping BIOS entry creation", efibootmgr had both "Unraid Internal Boot" entries at the top of BootOrder. After a clean reboot from the GUI, rc.6 transferred /boot and the server now boots from the SSDs:

- /boot = flash/boot (zfs)

- zpool status flash: mirror ONLINE, both partitions ONLINE, 0 errors

- BootCurrent = Internal Boot entry

The question

Main shows "Internal Boot" with "Boot Partition (ONLINE)" and "Data Partition" (devices shown as new). In disks.ini, boot/boot2 are DISK_NEW (type Cache, fsType auto) while flash/flash2 are type Boot, DISK_OK. config/pools/boot.cfg has diskBootSize="0", diskId="" and diskIdSlot set to the two SSDs.

Every array start shows:

Unmountable disk present: Boot (sdab), Boot 2 (sdac) - Format will create a file system on all unmountable disks.

The "data partitions" are partition 4 on each SSD, about 1.8 MB - the leftover behind the boot area. I have NOT formatted.

1. Is this the expected state for a dedicated boot pool, or did the timeout leave the pool half-registered ("not boot eligible", empty diskId)?

2. Is there a supported way to clear the data-partition part so the format prompt goes away, without redoing the whole migration?

3. Does the 7.3.0 known issue about a dual-device boot pool with an unformatted data partition not completing device removal apply here, i.e. would replacing a failed boot SSD be a problem in this state?

Happy to provide diagnostics if needed.

Thanks!

Thanks for the detailed write-up, and good call not formatting.

1. This is not the expected end state. The partitioning, ZFS mirror and bootloader install all completed, which is why the server boots fine from the SSDs. But the last step of the setup, where Unraid registers the pool as a dedicated boot pool, failed. That is the cmdMakeBootable: not boot eligible line followed by the HTTP 500. A correctly finished dedicated boot pool shows on Main as "Internal Boot(Dedicated)" with only the boot partition, and there is no data partition to format. Yours shows "Internal Boot" with a Data Partition row. That suggests the pool ended up registered as boot + data, so the ~1.8 MB leftover partition 4 is treated as an unformatted data pool.

2. Don't format it. That would not fix the registration, and there is no usable space there anyway. In the meantime you can start the array normally. Just leave the format checkbox unticked. Before suggesting any change, I'd like to see the diagnostics (Tools > Diagnostics, after a normal boot; the array can be stopped). They will show exactly how the pool is registered. If you still have the syslog from the boot where you ran the wizard, that would help too.

3. Yes, that known issue is about this kind of state: a dual-device boot pool with an unformatted data partition. So I would rather get this sorted out before you ever need to replace one of the SSDs. Your boot mirror itself is healthy, so this does not put the current boot at risk.

Please post the diagnostics and we'll go from there.

Thanks JorgeB! I've sent you the diagnostics via PM 🙂