Hi Cyril!
On Sat, Jul 11, 2020 at 11:15:03PM +0200, Cyril Brulebois wrote:
> Hi Nicholas,
>
> Nicholas D Steeves (2020-07-10):
> > My plan is thus:
> >
> > 1. After we have installation to subvolumes, add subvolume listing support
> > to the rescue cd. This has the side-effect of being able to test "boot
> > environments" from the rescue cd.
> > 2. Activate boot environment support via grub-btrfs (#941627).
> > 3. Long-term: add debian-installer support for user-configurable subvolume
> > layout like Fedora and OpenSUSE have. Ideally I'd like to work on this as
> > part of a btrfs-enablement team
> >
> > Thanks,
> > Nicholas
> >
> > CCing Cyril for Kibi ACK :-)
>
> Please don't block on me. I'd assume you have much more experience with
> btrfs than I do, and I'm not really concerned about possible breakages
> with that particular filesystem (upon checking, it appears it doesn't
> even appear in the installation guide for some reason).
>
Sorry for the delay replying, this was a case of "gmail silently
ate/bounced your email" :-(
Re: installation-guide: If I remember correctly someone recommended
striking mention of btrfs from the installation guide because they
believed the fs was RC buggy. Given that Fedora 33 is allegedly going
to default to btrfs (with subvolumes for rootfs and home), for real
this time, I think it's clear that this is no longer a just
assessment.
Thank you, I appreciate your vote of confidence! :-) The most significant
potential issue I'm aware of is if Debian begins to default to using a
persistent journald. Upstream journald defaults to using "chattr +C"
(eg: nocow, no checksums, no protection), and there was a historical
bug where this combination caused problems, but given the Fedora news
I think it's fair to suppose that this potential issue has been
resolved upstream.
I'll wait a month to Andrew Hayzen a chance to test, then retest,
then merge.
Oh, and there is a fourth long-long-term objective: enabling advanced
btrfs features such as btrfs-native raid profiles and compression in
the installer. This one is without a shadow of a doubt post-bullseye,
because util-linux and coreutils aren't yet sufficiently aware of
btrfs' freespace accounting peculiarities when using these features
(said another way, btrfs doesn't export expected info). I'm not
blaming one group or the other, just saying I don't believe these
features are not yet "Debian stable" quality in terms of user
experience. I expect it will be ready for bookworm!
Ideally I hope Adam Borowski, Hideki Yamane, and maybe Andrew Hayzen
help found the future btrfs-enablement team :-)
Cheers,
Nicholas
signature.asc
Description: PGP signature
