From Fedora Project Wiki

Revision as of 17:37, 20 August 2019 by Adamwill (talk | contribs) (update sample config and log output)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Associated release criterion
This test case is associated with the Fedora_41_Beta_Release_Criteria#direct-kernel-boot release criterion. If you are doing release validation testing, a failure of this test case may be a breach of that release criterion. If so, please file a bug and nominate it as blocking the appropriate milestone, using the blocker bug nomination page.


Description

This is to verify that it is possible to boot Anaconda and install the system by direct kernel+initrd boot. That can be achieved either by using PXE boot or by booting the kernel directly in a virtual machine.

Setup

  1. You will need vmlinuz and initrd.img files. Those are located in the (arch)/os/images/pxeboot directory of the tree for the Fedora compose you wish to test.
  2. You will need a remote location containing Anaconda's LiveOS/ directory (containing the installer) and optionally also a package repository. Development composes are usually available at dl.fedoraproject.org/pub/alt/stage/ (doesn't contain package repository). If you have a local mirror, you can make it accessible over any protocol supported by inst.repo boot option.
  3. Option 1: Set up a PXE server.
    • This is not an easy task and requires some administrator knowledge. You can read the appropriate section in the Installation guide for Fedora 40.
    • The custom pxelinux config file can look like this:
      default vesamenu.c32
      prompt 1
      timeout 600
      
      label linux
      menu label ^Install Fedora
      menu default
      kernel fedora/vmlinuz
      append initrd=fedora/initrd.img inst.repo=https://dl.fedoraproject.org/pub/alt/stage/41_Beta-1.1/Server/x86_64/os/
  4. Option 2: Boot kernel directly in a virtual machine.
    • Prepare a virtual machine that can boot kernel+initrd pair directly, e.g. using virt-manager (virt-manager has a Direct kernel boot field where you can specify the kernel, initrd and any boot options).

How to test

  1. Boot the system via PXE, or using a virtual machine with appropriate inst.repo boot argument (it has to point to the same compose you used for retrieving vmlinuz and initrd.img).
  2. Proceed with installation.

Expected Results

  1. The system boots (using PXE or direct kernel boot in a virtual machine) and it downloads anaconda installer from the specified remote location.
  2. The installer starts correctly.
  3. If the remote location contains a DNF repository, that repository is used for installation. This can be checked by examining the /tmp/packaging.log file. Example output:
    09:35:25,625 INF packaging: added repo: 'anaconda' - https://dl.fedoraproject.org/pub/alt/stage/31_Beta-1.1/Server/x86_64/os
    09:35:25,647 DBG dnf: repo: downloading from remote: anaconda
    09:35:27,161 INF packaging: enabled repo: '' - ['https://dl.fedoraproject.org/pub/alt/stage/31_Beta-1.1/Server/x86_64/os'] and got repomd
    09:35:27,163 DBG dnf: repo: using cache for: anaconda
    09:35:27,163 DBG packaging: repo anaconda: _sync_metadata success from ['https://dl.fedoraproject.org/pub/alt/stage/31_Beta-1.1/Server/x86_64/os']
    09:35:27,165 DBG dnf: repo: using cache for: anaconda
    05:35:50,985 DBG dnf: anaconda: using metadata from Tue 20 Aug 2019 02:34:29 AM EDT.
    
  4. The installation completes and the new system initiates boot properly