Report: Sun Open Storage

[Mark Schouten]

On Wed, 2008-11-19 at 06:27 +0900, Onno Benschop wrote:
> I'm not saying that their solution is crap, I'm saying that they're
> telling me one thing and offering me another. They're telling me the
> machine is a real server, "it's running Open Solaris was the mantra",
> but when I actually want to use it as a server (which personally I think
> would be an excellent idea - and I'm interested to hear comment on
> this), I void my support contract which makes no sense to me at all.

So we're having a discussion on the Ubuntu server mailinglist, about how
you misinterpreted a marketing line from Sun?

> Ironically, the VMware issue came up and I suggested to the Sun engineer
> in front of me at the time that if they actually had real VMware
> certification, why didn't they offer to run appliances on the machine,
> and amend the support contract to include something like this: "If your
> problem is caused by your running VMware appliance, Sun support will be
> unable to assist you, however, if when the appliance is stopped and the
> issue persists, you'll receive full Sun support." - but I suspect that
> it will be some time before we see something like that :-)

^^ Still missing the point of the box...

> Which reminds me, there was no discussion about what happens to their
> system during upgrade. There is a roll-back for upgrades, but there was
> no discussion about what happens during the upgrade and no reference to
> interoperability between clustered solutions either (other than to say
> that interoperability was extremely closely tied to firmware versions
> and OS versions), so there is no information on if two or more clustered
> devices can run together with different versions, so you can reboot one
> after an upgrade without turning off the cluster - I suspect "that's in
> a future release".
> 
> A final Ubuntu-server thought, the roll-back idea seemed like a really
> cool thing that we could implement with a snap-shot. That is, do a
> system-snap-shot before any upgrades leaving the ability to roll-back a
> system if the upgrade had issues - of course little things like incoming
> mail and database queries might be a problem, but if we deal with that
> by separating the OS from the data (hmm, where did I hear that before
> :), then we might have ourselves a feature that I know I'd use. Nothing
> like doing an upgrade at midnight, having it fail and spending the next
> 8 hours fixing it :)

A dist-upgrade has never taken me more than two hours. Falling back to a
snapshot is nice if you don't have (or take) the time to do your
upgrade. But I'd rather fix the problem I see when upgrading than going
back a snapshot and start of over again (running into the same issues).

-- 
Mark Schouten <mark at prevented.net>