Re: [Apcupsd-users] [PATCH] Conf: new SELFTESTGRACEPERIOD option to prevent self test shutdowns

[Jan P. <po...@fn...>]

Hello Pavel,

trying to follow in-line...

Pavel Boček píše v Pá 20. 10. 2023 v 10:34 +0200:
> Maybe I've missed something, but I'm unable to find anything (which
> could just mean the messages ended in SPAM folder aigain :D) in the
> list, so you have likely not discussed this before here, correct?

correct ...

> So the problem is, as I understand the comments, the UPS starts self
> testing when it already has somewhat discharged accumuators, which
> results in LOWBATT signal, starting the procedure of apcupsd shutting
> the machine down, right?

... and on this account as well

Luckily for us, apcupsd as a state machine is free to choose to ignore
that particular "fickle subtlety", which is exacly what I am trying to
achieve in the patch, on an opt-in basis.

> That once again appears to me like trying to solve something much
> later what should have been solved before actually buying anything.
> This series is known for decades as not having the HW AND SW (or
> rather, FW) equipment for what you would have expected of high-end
> hardware, despite APCs pricing and marketing could be somewhat
> overstating the actual position of this device in their product
> portfolio. Not that the CPU inside would not be able to handle that,
> but because APC (Schneider) made it so.

Not arguing anything like that, but please consider (and this also
responds to some later parts of your response) that this device was
purchased dozen years ago, still works satisfactorily (maybe even
perfectly with the original SW for non-foss platforms, cannot tell)
to this day with new batteries in it, and the only annoying bit
I witnessed first hand is what the patch tries to resolve.

Sometimes it is simply not justifiable (economically nor ecologically,
perhaps) to rush buying "etalons of quality" (dumping still working
old stuff) in given circumstances only to demonstrate a "purchaser's
voting power".  Would there be no other way around than to buy a new
thing, sure, the decision shall be made after evaluating pros and cons
carefully beforehand, so that uninteroperability and/or crappiness
of HW is, ideally, pushed out of the market early on.

Frequently though, one needs to deal with what's available, and last
thing I wanted with the patch proposal was fall into such armchair
philosophical debates.  I'd rather want to discuss the technical
aspect of things.

> I am quite possitive that Smart-UPS series clearly does not show such
> stupid behaviour, as being their business/entry-level-mission-critical
> piece of HW, it is an etalon (in the means of "standard") of quality.
> So, amongst others, the self- test function clearly has condition to
> start only at 100% charge, when anything happens, the tests stops
> immediatelly and the device starts indicating problem (=showing
> REPLACE BATTERY and making noise every 10 minutes or so), also the
> unit measures same parameters as in self-testing during EACH power
> outtage, and generaly it is "smarter" (=better HW equipped and better
> programmed).
> 
> In short, ppl often come here in hopes of solving problems, they would
> never have to be dealing with, haven't they chosen inferior,
> cheaped-out product in the first place. I can see from the address
> that you work in one of the hospitals, so I know well how shopping
> works in such instituions, when the first, second and third top
> criterias are price only. On the other hand, there are much better
> possibilities in terms of shopping for "critical" equipment in
> hospitals than at, say, state/public operated universities or hotels
> (=imagine 10times worse possibilities than you have in your
> organisation).

Yeah, and to make it clear, that decision process predates my tenure
by a looong shot and I have no idea about the models available at
that time nor about pricing of the alternatives
... late nineties/early 2000s it seems :-)

> I mean, yeah, having such patch is great, I intend this rant more like
> a warning for ppl to do their research beforehand, and especially
> definitelly do NOT buy inferior devices in hopes that you can mitigate
> the problems later on the OS or the user side in general. Which is
> exactly what many ppl will do anyways after (if) Ted implements this
> :D

As mentioned, we need to find some common middle ground, which for
me is:

- new procurement: choose wisely, giving the best feedback to
  the market one can ever make

- need to deal with less-than-optimal stuff at one's disposal,
  regardless of why:
  let's accept/solicit volunteered best effort to satisfy the
  affected userbase, as long as it won't cast a negative impact
  (that being said: is a new configuration item bad for overall
  cognitive load on the users? is really the only thing coming
  to mind), and while refraining from false proclamations/promising
  wonders

I think a poster child of that is Linux itself, just look at
how many quirks and workarounds are there to make even the users
with questionable HW happy.  It would hardly be possible if the
prevailing attitude was a nonconstructive naysay.

> UPS should be, in my view, highly autonomous devices, correctly
> reporting either its operating states, or controlling conencted device
> directly for it to (when it has to) shut down gracefully, without any
> secondary damage.

Putting electricity-inflicted damage aside as an extreme and unlikely
case, the overall robustness required depends on the criticality of
the deployment, and there sure is a load of nice-to-have-power-backup
as an extra value added, not a vital requirement, just as here with
my workstation.  Just this little nit that gets on my nerves ...
when fixed in apcupsd, I am good.

On that front, the patch is deployed locally, as mentioned, will report
back, but if anyone spots any logical gaps or anything, as well as
patch inclusion blockers, please let me know.

-- Jan Pokorny (poki)