#3063 Should we remove Systemd-boot support from Anaconda
Closed: Rejected 2 years ago by zbyszek. Opened 2 years ago by jkonecny.

Systemd boot is accepted change to Fedora:
https://fedoraproject.org/wiki/Changes/cleanup_systemd_install

And was implemented to Anaconda by this PR:
https://github.com/rhinstaller/anaconda/pull/4368

However, we have this bug recently filed on Anaconda:
https://bugzilla.redhat.com/show_bug.cgi?id=2234638

When I was looking on the bug I found that the issue is raised because of missing package in Fedora (see comment 5 on BZ). I've set needinfo on author of the change but so far no reaction.

I'm contacting FESCO to decide about the future of this feature. Should we reverse the commit from Anaconda to avoid crashing or are you able to somehow solve the missing package in Fedora 39?


Hi,

I'm alive, but as can be seen in the package history there isn't much movement on the package review. Although, maybe its me, let me see if the branch creation works.

So, I guess I missed the review+ last week on the package since it was molding for so long.

https://bodhi.fedoraproject.org/updates/FEDORA-2023-3b6fc9579c

So it should work with the everything images, but probably won't out of the box for workstation/etc without the missing comps change to pull it into those images.

Workstation don't need comps change but fedora-kickstart change

https://pagure.io/fedora-kickstarts/tree/main

or better just add this package as Anaconda dependency so it's dragged everywhere:

https://github.com/rhinstaller/anaconda/blob/master/anaconda.spec.in

Feel free to propose PR when it's ready.

It can't be pulled in via a dependency like that (AFAIK) because it conflicts with grubby. The comps change is the right place for it, because it ends up in the right repo groups, rather than actually being installed.

Sounds correct. Thanks for explanation.

Another 10 days with no progress… This does not look good.

In the review ticket, I wrote (https://bugzilla.redhat.com/show_bug.cgi?id=2134972#c12):

Also, the package has Conflicts:grubby, and grubby is depended-on (transitively) by many packages, so this would prevent this package from being installed, or grubby from being installed. It's much better to avoid Conflicts, and instead allow packages to be installed together, and DTRT depending on the state of the system.

It seems like my initial assessment was correct ;) This design seems rather problematic. But if you want to keep it, OK. But it needs to be made to work quickly. We're already after beta.

Does anyone know if anything changed here?

Metadata Update from @zbyszek:
- Issue tagged with: meeting

2 years ago

I'm confused what you mean by no progress. There is an open PR for anaconda to provide a popup on live media. https://github.com/rhinstaller/anaconda/pull/5172

There have also been various other fixes for the btrfs issue on the everything media (https://github.com/rhinstaller/anaconda/pull/5194), the kickstarts exclusions have been merged, which should allow the comps change to merge which will fix the server image not having the sdubby package.

I'm looking at ostree/install issues now.

And of course some of this is being found by people other than me now that it doesn't take half a day to setup a configuration that can test it.

And it does work, it just needs to avoid the assorted fixes here/there that are the result of it largely being ignored for a year and the difficulty of testing install media, without actual install media.

Its possible to install a system-boot system with both the server and everything media by either using the network mirrors to acquire the sdubby package (for the server dvd) or selecting LVM or XFS, or simply adding the rootfs= options for btrfs on the everything media.

And just to add to this, none of it is particularly hard to fix, the live media could be provided with a conversion script to convert from grub to systemd, its just not been part of my personal targets as I've largely been focused on just normal installs.

This was discussed during today's FESCo meeting:
REJECTED (i.e. the support for sd-boot is not removed) (0, 0, -6)

Metadata Update from @zbyszek:
- Issue close_status updated to: Rejected
- Issue status updated to: Closed (was: Open)

2 years ago

Metadata Update from @zbyszek:
- Issue untagged with: meeting

2 years ago

Log in to comment on this ticket.

Metadata