Cannot run “proxmox-boot-tool format” because of “Invalid partition data!” message

[Wouter Samaey]

I too got stuck with an unbootable legacy ZFS setup and am just now trying to switch to ESP using proxmox-boot-tool.
I don’t have any partitions on hand that could be used, but I do have a ZFS cache disk that I have removed, partitioned and would now like to use as an ESP. I’ll re-add as cache later once I can boot again.

So far, the above has worked using the Proxmox installer rescue console.

Where I get stuck is the “format” step. I always run into the “ Invalid partition data!” message and I don’t have a clue what could be wrong. I used fdisk to create a new partition of 512MB and there’s nothing else on the disk. The disk is an NVMe SSD.

Any ideas?

 

[Kingneutron]

Did you set EFI partition type to EF00 in gdisk ?

 

[Wouter Samaey]

thank @Kingneutron . That was it.
Iwas now able to continue with proxmox-boot-tool, but when finished and rebooting, I ended up in Busybox without any errors. I believe ZFS rpool was not imported somehow? I’m seeing on this forum some peope have issues with /etc/default/grub config, but mine looks okay.

Any ideas what it could be? Maybe i need to reset the mountpoint? That’s something I didn’t do…

 

[Kingneutron]

Hmm, if you ended up in busybox try ' zpool import -a -f -d /dev/disk/by-id '

Although if it can't boot ultimately, your best bet is to copy all the configs off ( /etc/* ) from a rescue environment (you can do this with tar to an sshfs mount, local separate disk, or NAS) and just reinstall to another zfs mirror. Then restore from backup. Not sure if you have anything besides the OS on rpool, but it's a good idea to separate OS+data and have VMs on a separate pool for DR situations like this.

https://github.com/nchevsky/systemrescue-zfs/releases

 

[Wouter Samaey]

Thanks for your help so far. I was able to mount the zfs rpool (i only have the 1 zpool and everything is in there), but after mounting I’m not sure what should be next. Leaving Busybox with CTRL+D isn’t enough as a new error appears.



I’m a bit out of my league here, so thanks a million

 

[Kingneutron]

Go back to my last post. You're going to have to copy /etc and whatever else off, and reinstall

 

[Wouter Samaey]

Hey @Kingneutron . I took your advice. Bought 2 small SSDs and installed Proxmox fresh on these 2 disks using ZFS Raid1. These will be for OS only.
Proxmox is now booting, and the previous zpool is available (i did rename it so not to conflict with the new rpool).

Is there a guide on next steps?

I’m thinking about:
- Recoving my previous vms, but making sure they run on the big zpool and not on the small SSDs (likely default configured I imagine)
- Installing docker and setting it’s locations right (I have a few Docker containers that I use even though Proxmox doesnt have that natively)
- Making sure new VMs and containers are stored on the big zpool
- Manually moving over some /etc config, not sure what I need yet
- Anything else

 

[Kingneutron]

- Manually moving over some /etc config, not sure what I need yet

Look in /etc/pve/nodes for VM and CTR config files