[SOLVED] suddenly, grub fails to boot PBS

[giovvv]

Today, I started my Proxmox Backup server, and grub fails to boot.

The server is a repurposed old machine with two 4TB HDD with ZFS RAID 1.

This kind of failure happened to me in the past with PVE, when I was trying to have ZFS disks connected to a Dell PERC controller (an unhealthy idea that I later abandoned). But this time it is not that case: this is a very simple desktop machine with no strange controller and the HDDs directly connected to the motherboard.

Enclosed are the snapshots of GRUB (set debug=all).

What can I do now?

 

[giovvv]

For anyone that could encounter the same problem: turns out the problem was that I had created a zfs dataset (in addition to the ones created by the PBS installer) without specifying the dnodesize=legacy option. When grub encounters a dataset with dnodesize=auto (the default), it fails to boot.

I solved it this way, rather quickly and with no data loss:

1. Removed the disks from the server and putting them in an USB enclosure.
2. Connected the enclosure to another zfs-capable machine; imported the zfs pool there.
3. Found out the problematic dataset with zfs get -r dnodesize rpool
4. Moved the contents of the dataset to a brand new one, this time created with dnodesize=legacy
5. Destroyed the offending dataset
6. Exported the zfs pool
7. Put back the disks in the PBS server, which now is able to boot as before.

For more details see this article from which I got the general idea.

 

[giovvv]

Not just that. Grub also fails to boot if any encrypted dataset is present in the pool. See here