[Nut-upsdev] Asking hard questions about the NUT architecture

[Eric S. Raymond]

Rob MacGregor <rob.macgregor at gmail.com>:
> Now, generally there's nothing you can do, even if you're right there
> (unless it's just a trip switch going).  However having that 20
> minutes (or however long you've got) to notify people, start switching
> key services to other systems, or whatever your processes say - that's
> a good thing.  Finding out that you've lost power by having the system
> monitoring screen turn red - that's not a good feeling.

Fair point.  This is a good case for remote notifications.

> ] I'm not (just) being snarky.  You seem to be kind of wobbling back
> ] and forth between use cases for UPS-controlled shutdown and
> ] remote alerts.  I agree that both actually have use cases, but
> ] the cases for one don't support the other very well.
> 
> Why, in your opinion, don't they?  I've found them useful, more than
> once.  Having time to gracefully warn people to save work, shutdown
> services and start them elsewhere is a good thing.

This is still a case for remote notifications.  But I think the 
case for UPS-controlled shutdown is orthogonal -- it depends on factual
questions about how well the system lands when you cut power to it.

That doesn't have much to do with the case for remote notifications,
because if the machine doesn't land well you want UPS-controlled
shutdown whether or not you're raising alarms elsewhere; conversely,
if your procedures need remote alarms, that's a requirement whether
or not the box lands well.

> More seriously, what if I have a need/desire to manage multiple UPS
> with a single host, and some are USB and some are serial?  With your
> plan that would require multiple systems (at least virtual ones, if
> not physical).
   :
> Also known as, upon what do you base your assumptions about the
> percentages of users who'd be using this new Linux only/USB only
> version?

Size of the potential userbases.  For every large-system sysadmin who
actually needs a setup like that, I would be astonished if there were
fewer than a hundred single-UPS/single-system setups out there.  Just
looking at the piles of consumer-grade USB-UPS boxes at Computer
Center told me that --  the store expects to sell those in *volume*.
 
> > What does BSD use as a power-fail signal, then?
> 
> That I can tell, it doesn't.  However the amount of time between the
> power going out and the system stopping isn't really long enough to do
> anything useful except flush the disk cache, maybe.

That'd be a heckuva start. I'm actually kind of shocked to learn that
BSD has such poor hardening.  I guess I still had some lingering belief
in the BSD propaganda about their kernel being better architected. 
-- 
		<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>