GSoC Project Involving the reimplementation of beadm(1)

[Slawa Olhovchenkov]

On Tue, Feb 21, 2017 at 10:20:00PM +0900, Tomoya Tabuchi wrote:

> On Mon, Feb 20, 2017 at 04:49:10PM +0300, Slawa Olhovchenkov wrote:
> > On Mon, Feb 20, 2017 at 10:15:09PM +0900, Tomoya Tabuchi wrote:
> > 
> > > Hello,
> > > 
> > > I am interested in doing a GSoC project this year with the idea "Write a
> > > new boot environment manager" on the ideas list.
> > > (https://wiki.freebsd.org/SummerOfCodeIdeas#Write_a_new_boot_environment_manager)
> > > 
> > > I would like to ask a few questions involving this.
> > > First, is there a particular reason why this project is listed in the
> > > ideas list? Aside the fact the current implementation in sh is rather
> > > complicated, I was unable to come up with a reason to justify the
> > > reimplementation.
> > > 
> > > Second, is making the new implmentation of beadm(1) platform independent
> > > and promoting it across the various OpenZFS implmentation / communities
> > > as some sort of "standard" implmentation a good idea, or is it
> > > over-zealous / outside of the project scope / intrusive to other
> > > projects.
> > > 
> > > As for a late self introduction, my name is Tomoya Tabuchi, and I am a
> > > undergraduate student at Doshisha University in Japan. I will start my
> > > third year in university in April.
> > 
> > Don't know about link above. For me, current beadm have some leaks:
> > 
> > 1. Don't check cosistency before applay:
> >   I am try to enable beadm on 10.1 install and switch to 11.0.
> >   fail.
> That is interesting. I'll try and see if I can reproduce that, and
> observe what's going on.

This is may be involved by outdated /etc/rc.d/zfs from 10.1 or by
different ZFS layout in 10.1.

(how to update by beadm:

beadm create NEW
beadm mount NEW /mnt
cd /mnt
find -x . -flags +schg | xargs chflags noschg
unpack new binaries
cd /
beadm umount NEW
beadm activate NEW

AFTER REBOOT:
etcupdate -t etcupdate.tar

i.e. new version boot w/ old /etc/rc.d

In may case I am see multiple mounted datesets w/ same mount points
and lack of /dev

For 10.2 and later -- OK.

> > 2. Need to control what put under beadm.
> Does this mean to hold back on feature bloat, or to distinguish between
> ZFS clones created by beadm and ones that were not? If you mean the
> latter, I'll take a look at the current behaviour when manually created ZFS
> clones are involved.

Yes, but this is not intuitive.
I am about more friendly from beadm, i.e. for example:

beadm showcontrol -- show path beadmed
beadm takecontrol -- do some path beadmed
beadm releasecontrol -- reverse

yes, in many cases need addtional confiramtions for split datasets and
move some files between. and warning about inconsistency old.