Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 15 Jan 2020 21:19:51 -0800
From:      Mark Millard <marklmi@yahoo.com>
To:        Ed Maste <emaste@freebsd.org>, freebsd-arch <freebsd-arch@freebsd.org>
Subject:   Re: svn commit: r356758 - in head/usr.sbin/bsdinstall: . scripts
Message-ID:  <2C0CC3EC-343D-4068-9E05-856A66FF175C@yahoo.com>
References:  <2C0CC3EC-343D-4068-9E05-856A66FF175C.ref@yahoo.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Ed Maste emaste at freebsd.org wrote on
Thu Jan 16 02:14:58 UTC 2020 :

> On Wed, 15 Jan 2020 at 17:55, Warner Losh <imp at bsdimp.com> wrote:
> >
> > It's not i386 vs amd64 that's the issue. It's the total amount of =
RAM. ZFS
> > is perfectly happy with i386 installs with enough RAM. We'd catch =
most
> > sub-optimal cases if we'd offer it only for systems with > ~2-4GB.
>=20
> Very good point, defaulting to UFS for <4GB and ZFS for >=3D4GB sounds
> decent to me.

I'd pick to avoid Rock64, RPi4, etc. in 4 GiByte
configurations ending up with ZFS by default, suggesting:
UFS for <=3D 4 GiByte and ZFS for > 4 GiByte if 4 GiByte
is to the at the boundary.

What fraction of the <=3D 4GiByte systems are such
contexts?

(I'd probably bundle <=3D 4 GiByte PowerPCs with
such Arm contexts but they may not add much.)

It also might be that the snapshots and release
builds might not use the same criteria as the installer
defaults. (Only applies to those that end up with
such snapshot and release builds. But the RPi4 would
seem likely to get such builds someday.) It might
be that snapshots and releases would consider how
likely uSD cards for the known target cases. (I do
not know what the UFS vs. ZFS tradeoffs are for
such media issues. This might be an area were
PowerPCs with <=3D 4 GiByte is not a match.)

A consideration for placing the boundary might be how
much extra support configuring might happen on the
lists. An example is those trying buildworld buildkernel
on these sorts of physically-small systems. Likely
such folks will try -j4 just to try to cut the time
involved. I'm not sure how well this would go with
a default ZFS configuration. (And llvm materials are
likely to continue to grow for what it takes to
build them, so what might be marginal now later
could be a problem.)

Which is easier to support on the lists for these
sort of contexts, if they are a notable fraction
of the <=3D 4 GiByte usage? UFS? ZFS?

=3D=3D=3D
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?2C0CC3EC-343D-4068-9E05-856A66FF175C>