Can't access requirements.cgroup: v1 lxc container images

[kenhys]

Select Topic Area

Bug

Body

Problem

On GitHub Actions, it can execute a lxc container image.
But recently I've noticed that lxc container image which
requires CGroupV1 host environment can't accessible anymore.

On ubuntu-20.04 host, it is expected to be able to launch cgroup v1 lxc container image.

Actual

Previously, the following lxc container images were launched correctly on ubuntu-20.04.

• images:centos/7
• images:amazonlinux/2
• images:amazonlinux/2023

But these container images can't be launched anymore.

It is strange behavior that from CGroupV2 host (e.g runs-on: ubuntu-latest), above images are visible. (e.g. lxc image list images:)
(Surely they can't be launched at all, so it is useless)

From CGroupV1 host, list of lxc container images is empty. (unexpected result, they should be listed)

  To start your first container, try: lxc launch ubuntu:20.04
  Or for a virtual machine: lxc launch ubuntu:20.04 --vm
  
  +-------+-------------+--------+-------------+--------------+------+------+-------------+
  | ALIAS | FINGERPRINT | PUBLIC | DESCRIPTION | ARCHITECTURE | TYPE | SIZE | UPLOAD DATE |
  +-------+-------------+--------+-------------+--------------+------+------+-------------+
  + echo ::endgroup::

How to reproduce

• Use runs-on: ubuntu-20.04 (CGroupV1 host)
• Try to launch cgroup v1 image

It raises the following error:

Error: Failed getting image: The requested image couldn't be found
  Error: Process completed with exit code 1.

Here is the sample .github/workflows/linux.yml:

name: Check available lxc images
on:                        
  push:                           
    branches:                                                         
      - main               
  pull_request:                                                       
jobs:                            
  test:                                                               
    name: Test                                                                                                                              
    runs-on: ubuntu-20.04
    strategy:                                                         
      fail-fast: false                                                
      matrix:                                                         
        label:
          - CentOS 7
        include:                                                      
          - label: CentOS 7
            image: images:centos/7
    steps:    
      - uses: actions/checkout@master
      - uses: canonical/setup-lxd@v0.1.1
      - name: ${{ matrix.label }}
        run: ./launch.sh ${{ matrix.image }}

Here is the sample launch.sh:

#!/bin/bash

set -eux

image=$1

echo "::group::Run test: launch $image"
lxc launch $image target
sleep 5
echo "::endgroup::"

[kenhys]

I've overlooked the following articles.

• https://discuss.linuxcontainers.org/t/important-notice-for-lxd-users-image-server/18479
• https://discourse.ubuntu.com/t/new-lxd-image-server-available-images-lxd-canonical-com/43824

So, if you using runs-on: ubuntu-20.04 as a runner, you should switch remote image server for cgroup v1 image.

 - uses: canonical/setup-lxd@v0.1.1
 - name: Update image server
   run: |
       lxc remote remove images
       lxc remote add images https://images.lxd.canonical.com --protocol=simplestreams

On ubuntu-22.04 or later, it seems that already remote image server was switched to images.lxd.canonical.com by default.