[oi-dev] OpenIndiana and illumos

[Garrett D'Amore]

On Thu, 2010-11-18 at 15:57 +0900, BM wrote:

> 
> > But I
> > also stand by my remarks; using a first version development release
> > product with no commercial support and not real software engineers
> > "supporting it" is nuts.
> 
> BTW, it is just 134+ rebranded, essentially. 

Nexenta's kernel is quite a bit more than "just 134+".

> But taking whole picture,
> there is no free Solaris around then. So what is more stable and
> better supported for you: FreeBSD with wobbling ZFS via kernel mapping
> or snv 134?

Um... for production, I think if you want to have a "production" release
you need to look to older "stable" OpenSolaris (2009.06, which is now
17months old), or a non-free option.

See, you forgot to add something in your qualifications, which is you
want something very recent. 

> The bottom line is: personally I am *not* against your points and
> waiting when OI can finally move towards completely unstable
> nonproduction Illumos :-), but just... hey, don't beat people with a
> baseball bat. Be nice, specially because you are *leader*.

I've been trying to be "nice".  But I am concerned by what I perceive to
be a lack of progress on critical goals moving forward, and myopic
vision on a release that at this point is totally pointless (OI148).

> 
> We have two problems, basically:
> 1. People at *production*, who wants ZFS (and don't wants Nexenta or
> Snorcle's "up to eleven").

We should not ever have had people running OI or illumos in production
yet.  That was a huge mistake.  IMO, those people took the risks with
the warnings, and I don't think we should constrain our development
efforts in order to make life easy for people who insist on putting
bleeding edge bits into production. 

> 2. Moving to Illumos is a must.

Yes!

> 
> How would you suggest to address BOTH these issues in a right way?

I don't.  I think the first item should never happened, and worrying
about people who put something that clearly should *not* have been in
production (it was called a development release after all) artificially
constrains the problem.

Also, I'll point out, they aren't *forced* to upgrade to whatever is
next.

We simply don't have enough resources to solve all the problems.  So I
would prefer to focus on the ones that are going to help us going
forward rather than looking at trying to solve artificial problems to
make a few "brave early adopters" happy.

> E.g. I understand your points about all the things up to SX11, but
> killing #1 would be also not the best idea. 

Nobody is saying "kill it".  They can keep running OI 147.  They can
upgrade to SX11. They can reinstall OpenSolaris 2009.06.  But I refuse
to waste any of my resources trying to help *those* guys run a
development release of software that is not even on a code branch that
leads to my ultimate product.

> I also believe that
> Alasdair relied on Joyent too much (probably hence entire OpenIndiana
> is all about), hoping they will contribute, while they are not really
> doing so.

I've been disappointed by the lack of contribution from Joyent, as well.
Hopefully that will improve in the future.

> 
> Therefore I think at OI we should re-think our primary goals and
> re-target efforts as well as still "support" of what we already done
> smoothly. Nexenta 4 will be probably a disaster as well: as far as I
> know, there will be no smooth upgrade, once it on Illumos, but a
> hurricane in a style "kill all and reinstall from scratch"? Or I am
> mistaken? Hope, NP3 user base won't be just killed, because it is not
> Illumos... :)

NCP 4 will indeed be a reinstall.  We'll have a way to preserve and
restore config data for commercial NexentaStor customers, I hope.

There is a great deal of evidence which suggests that commercial
customers of Operating Systems do not generally "upgrade" anyway.
Probably this has a lot to do with both wanting to "start
clean" (removing stale things, etc.) and a general distrust of operating
system upgrade technology as well (which quite honestly has been well
justified.)

Technically, there isn't an easy way we could have made a simple upgrade
process from NCP 3 to NCP 4; while the problem could have been solved,
it would have been quite expensive to do so, and probably would have
been error prone.  So we're not doing it.

The NCP 3 user base, such as a it is, receives no active support from
Nexenta apart from what we give to NCP as part of our efforts around
NexentaStor.  We won't be doing anything special for those non-paying
NCP users.  (At the end of the day, Nexenta is a business, and we will
direct our efforts where the is commercial justification.)

	- Garrett