Custom Image doesn't boot after install

[max06]

Good morning... again 😅

I'm currently building an elemental distribution based on ubuntu 24.04 - on your orange flavor. It took some time to get used to this toolkit, as well as some fixes in the orange dockerfile to make things work.

So far I have a working build pipeline throwing out container and iso oci images, and an accompanying channel image.
Said images, prepared by rancher, boot perfectly fine in my virtual machines.
Installation works and finishes without errors.

After reboot I get greeted by grub containing 2 items in its menu. And then it falls apart:

Booting the recovery works though.

I know you're not supporting foreign distributions, and I wouldn't even be (seriously) mad if you close this issue. Maybe just a little bit... for a while... 😄

If you are interested: Awesome!

Here's the channel manifest:

apiVersion: elemental.cattle.io/v1beta1
kind: ManagedOSVersionChannel
metadata:
  name: elemental-ubuntu-bare-metal
  namespace: fleet-default
spec:
  deleteNoLongerInSyncVersions: false
  options:
    image: ghcr.io/max06/elemental-ubuntu-channel:bare-metal
  syncInterval: 10m
  type: custom

The repo can be found here. It's mostly a mix of the orange Dockerfile and your elemental-repository for the config files. The pipeline uses the Makefile.

The whole thing is work in progress, some parts are missing. I just wanted to solve this issue before moving on. I just don't really have a clue at this moment.

Thank you very much!

[max06]

After some debugging:

Grub on suse and grub on debian behave differently.

Grub on suse is able to read the bootargs.cfg file without specifying a subvolume via cat (hd0,gpt4)/etc/elemental/bootargs.cfg
Grub on debian doesn't like that. (hd0,gpt4) is empty, (hd0,gpt4)/@/ shows a state.yaml and (hd0,gpt4)/@/.snapshots/1/snapshot/etc/elemental/bootargs.cfg leads to the golden lands....

Edit: Before I start building the suse grub2 with its 244 patches to make it work... can you add the current snapshot iteration to a variable similar to passive_snaps? Grubs scripting abilities are too limited to find that value on its own.

[max06]

Good morning, and happy new year!

I'm really not looking for any kind of priority or nagging or something like that - but a rough estimation would be awesome for me to plan ahead.

The hardware for my small test cluster will arrive next week, the big proof-of-concept setup is due for end of the month. It's fine if this issue stays longer, I just need to know so I can move on patching the ubuntu grub - I don't know enough go to fix the bug by myself.

---

Different topic: I'm not sure how your labels work - it looks like you categorized this issue as a bug with the orange release. I'd say this is a bug with the elemental toolkit, requiring the suse-built grub as a hard dependency -> affects all flavors except suse-related ones. Please correct me if I misunderstood.

[frelon]

Happy new year @max06!

I was debugging this issue and I think what is happening is that your elemental-toolkit version is missing this code which sets the active_snap grub variable from this PR: #2207

Also, your bootargs.cfg will need the new syntax with root_subpath:

set kernel=/${root_subpath}boot/vmlinuz
set initramfs=/${root_subpath}boot/initrd

We will release a new version soon, until then you will have to build your own toolkit container, sorry for the inconvenience!

[max06]

Good morning @frelon, and thank you for looking into this!

I was already wondering about the existence of active_snap in the source-code and thought it might be a relatively easy fix to add this to the bootenv - I'm kinda glad now I didn't try it 🙈

No worries - I'll give it a shot and report back 🙇🏼‍♂️