Electric Eel RC.2 Upgrade Exported boot-pool GUI Anomaly

@Caleb Summary: Electric Eel Installation Fails on Mirrored NVMe Drives with Residual ZFS Metadata

Description:
This ticket addresses an unresolved issue with the Electric Eel installer failing on mirrored NVMe drives that were previously used with TrueNAS SCALE. Despite related tickets (NAS-131900, NAS-132081, NAS-133006) being closed, this issue persists under specific conditions.

Steps to Reproduce:

1. Install TrueNAS SCALE (Electric Eel) on mirrored NVMe drives previously used with TrueNAS SCALE.

2. Observe that the installation appears to complete successfully.

3. Reboot as prompted, and the system drops into the initramfs shell instead of booting into the OS.

Observed Behavior:

• The initramfs shell suggests a potential resolution (e.g., force-importing ZFS pools). While this temporarily allows the system to boot, subsequent reboots return to the initramfs shell.

Workaround:

To resolve the issue, I used the following commands to manually clear the drives from the installer’s Shell option:

blkdiscard /dev/nvme0n1

blkdiscard /dev/nvme1n1

After running these commands, the installation completed without issue, and the system booted normally.

Expected Behavior:

The installer should fully clear residual ZFS metadata or RAID bits from drives, preventing these issues.

Request for Improvement:

1. Integrate blkdiscard (or a similar low-level cleanup tool) into the installer’s drive preparation process.

2. Provide clearer guidance or error messages for users encountering this issue.

Additional Question:

The Electric Eel installer no longer offers the option to configure a swap partition. Was this change intentional? If so, what was the rationale behind its removal?

Related Tickets:

• NAS-131900 (Closed Without Changes)

• NAS-132081 (Duplicate, Closed Without Changes)

• NAS-133006 (Duplicate, Closed Without Changes)

Please let me know if further information is required. I hope this feedback helps improve the installation process for future releases.

Thank you for the detailed explanation and for pinpointing the underlying issue. I appreciate the guidance, especially as this is a low-priority matter. It's good to know the changes in EE will prevent this in the future. I'll follow your advice on the backup and fresh install if needed.

Once you have freshly installed, you can upload the backup of your configuration and it will restore all you're configuration from before the install.

thanks for the find_import output. It shows the problem, zpool detects a zpool named boot-pool that was once importable but is no longer. This typically means that there are filesystem labels on existing drives in your system that have been formatted previously and are still being detected by zfs because the previous filesystem label is still in-tact.

We have fixed this moving forward starting in EE but upgrades have the potential of running into this issue. There really isn’t an easy solution once you’re into this scenario. The path forward is to save a backup of your configuration to your local PC. And then fresh install EE-RC.2 to the boot drives again. During the installation process we wipe all partitions and ensure any zfs filesystem labels are cleared out as well.

Done.