Post AhrnuTo3GsVBvf5z28 by siosm@floss.social
(DIR) More posts by siosm@floss.social
(DIR) Post #AhrnuSvSXqpjCLYNwO by jorge@hachyderm.io
2023-09-30T21:43:25Z
0 likes, 0 repeats
Thinking aloud, @siosm if one were to build silverblue/kinoite on top of CoreOS so that we're following a stable -> testing -> next stream... I've seen some old discussions.Has anyone tried it? Conceptually what would we need to do? I'm interesting in digging to find out. heh.
(DIR) Post #AhrnuTo3GsVBvf5z28 by siosm@floss.social
2023-09-30T22:15:54Z
0 likes, 0 repeats
@jorge The problem here isn't really in the how, it's in the organizational part and QA part. Once we have containers as first class artifacts we can start looking at replicating that model: https://gitlab.com/fedora/ostree/sig/-/issues/2
(DIR) Post #AhrnuUlFilrCtGnGJE by jorge@hachyderm.io
2023-09-30T22:27:10Z
0 likes, 0 repeats
@siosm Oh ok, so we should wait on this one then. 😀
(DIR) Post #AhrnuVggHFnJlNf7p2 by siosm@floss.social
2023-09-30T22:38:23Z
0 likes, 0 repeats
@jorge Also note that fully reproducing this model requires QA and new release processes that we currently don't have in Fedora. In the current Fedora CoreOS release model, someone needs to push a button every two weeks to make a release.
(DIR) Post #AhrnuWiqQh7SyNgMpk by jorge@hachyderm.io
2023-09-30T22:51:45Z
0 likes, 0 repeats
@siosm Yeah,I am wallpapering over the real question which is "would a continuous delivery apptoach work for Fedora?" I guess technically it's still F38 but not really. Feels like most of the major components are following this model but organized into F38, F37, etc. My interest in this is this cycle has had kernel regressions (from upstream, not Fedora), that affect AMD users. So I kinda want a tad slower, but we see the value in aggressive kernel updates too.
(DIR) Post #AhrnuXiAkgAy2aNLQO by vwbusguy@mastodon.online
2023-09-30T23:18:12Z
0 likes, 0 repeats
@jorge @siosm Theoretically, wouldn't you have that by being on the rawhide channel for Kinoite, Silverblue, etc?
(DIR) Post #AhrnuYTfu4AoPubHSy by jorge@hachyderm.io
2023-09-30T23:20:50Z
0 likes, 0 repeats
@vwbusguy @siosm Yeah but I can't go the other way. I can rawhide myself to newer kernels but I also don't have a realiable way to go back to 6.4.x in case of a regression.I can stay on an old image with 6.4.x, but then I don't get the rest of the updates, which is worse.
(DIR) Post #AhrnuZJmmJrD1Wytgu by sdgathman@chatter.llamarific.social
2023-10-01T00:28:46.720165Z
0 likes, 0 repeats
@jorge @vwbusguy @siosm I use btrfs snapshots to have multiple Fedora versions in /boot and root. Name your snapshot f38 or f39, and put that in the grub config for the kernel. I always do this when upgrading or trying out the beta. And even after switching, I keep a read-only snap of the older systems.I also do this with LVM when btrfs is not an option.This is with regular Fedora - not Silverblue or whatever. I wasn't clear whether that was a requirement from your comment.If it is just the kernel you want to try, just install it from rawhide, and there is automatically a fallback for several versions (number configurable in dnf.conf).
(DIR) Post #AhrnuZzyFTbL8MiaRc by jorge@hachyderm.io
2023-09-30T22:53:35Z
0 likes, 0 repeats
@siosm So I was wondering if three throttle settings: next, testing, and stable would give us a bit more granularity than the lockstep F37, F38, etc. Especially since the kernels get backported anyway so why deal with the numbers at all?
(DIR) Post #Ahrnua7PnnYXVSCX32 by sdgathman@chatter.llamarific.social
2023-10-01T00:30:19.133701Z
0 likes, 0 repeats
@jorge @siosm @vwbusguy I did an article for the btrfs method:https://fedoramagazine.org/make-use-of-btrfs-snapshots-to-upgrade-fedora-linux-with-easy-fallback/
(DIR) Post #AhrnuagrfzudHOmqIq by sdgathman@chatter.llamarific.social
2023-10-01T00:31:52.644581Z
0 likes, 0 repeats
@jorge @siosm @vwbusguy Here is an article for LVM:https://fedoramagazine.org/use-lvm-upgrade-fedora/
(DIR) Post #Ahrnud3yr1aidfcKbw by jorge@hachyderm.io
2023-09-30T22:55:03Z
0 likes, 0 repeats
@siosm Alternatively more kernel devs and QA would be the right answer, but that's a different and harder problem to solve.
(DIR) Post #AhrnugCbBRGeNGfkxc by jorge@hachyderm.io
2023-09-30T22:57:56Z
0 likes, 0 repeats
@siosm AMD vram downclocking and the intel wifi regression kinda makes me want to be on CoreOS's 6.4.15. I guess that's what drove me to wondering about this.