Robust home automation

I’m curious if truly robust design for HA system is feasible.

By robust HA I mean a system that can be left running unmanned for weeks or months, and can withstand and recover from typical residential problems such as power and internet outages, and - important - controller (Vera) failure.

Here’s how I see it:

  • controller is battery powered to withstand 24-48 hours power outages. Having Vera directly powered from battery is much more effective then doing that via switching wallwart plugged into UPS

  • given inevitable problems with hardware, redundancy is a must, so there should be a kind of clustering option: redundant controller with identical scheduling, capable of verifying that the primary controller is alive and well, if not (router crashed, dongle locked up) it should be able to take over. Even better - the second controller would verify that scheduling works as expected.

  • redundant internet connection, if primary is cable then there would be backup DSL or cellular…

Thinking along those lines, what if:

(a) Vera always uploaded her current config to FindVera.com, for deployment in the event of an emergency reset
(b) Users could be instructed to do a Port Forward through their home router, to Vera, which Vera would (under normal circumstances) ignore traffic on; but,
(c) If Vera should wake up from a crash, say, she’d listen in on that traffic, for a “Ping” from FindVera (who would know Vera hasn’t checked in lately); and,
(d) The “Ping” would contain a special “reboot yourself using this known-good config” sequence sent from FindVera.

No matter what “failsafe” solution we propose, it will be complex for no other reason than the multitude of ways things can “go wrong.” :-\

A simpler model might be that all initial configuration of any Vera goes through FindVera.com, presumably sending it’s serial# to FindVera and then “tying” it to an account.

This initial configuration step would only include the basics such as Creation of an Account (etc), “claiming” the serial# of your Vera, and any management of Software versions (in case the saleable unit has been sitting on the shelf and needs an upgrade)

Any followup config of the Device-Room-Scene stuff would occur on the Local Vera, with config “uploads” as Libra indicated.

Then it could be a very simple “box swap” should the Hardware ever die. Someone would have to be physically present for that part, but it could be made extremely simple if Vera did the “phone home” using it’s serial#, and support (or us) could “claim” it using our FindVera Accounts.

I think this will be more practical than any kind of tethered active-passive solution.

For me, I’m waiting for the OEM version(s) to come out, and I intend to by a 2nd purely for backup purposes. I’m also looking for a much more compact FF, with the potential to directly wire it into [local] Battery I have to avoid the UPS (as 325 indicated). This is how my alarm is done, and it works well and it’s compact. Technically, I could power Vera from the same battery…

Ultimately, I think if we can get down to “basic box swap” and “web/online claim” then we’ll cover the bases well.

Of course, being practical, when exactly was the last time your router died? I go through lots of DSL modems, but I have routers from the dawn of time, and these never seem to die (they just get functionally obsolete)

For Software-based failure, the device will mature over time. I recall early versions of OpenWRT itself (and others) and these have come a long way. Most of the problems that people have experienced seem to relate to ZWave components. Perhaps this is why mine is ok, as I have very little ZWave, and mostly custom interfaces.

I was going to make a joke about having a rotating carousel full of Vera units, which would automatically rotate and plug in a new Vera any time one stopped working. ;D

But, seriously … what if Mi Casa Verde began an easy-to-navigate accreditation program whereby techs - who can provide in-home setup and configuration for “regular” customers - would pay a small fee for (a) a certificate, (b) a listing at MCV, organized by state/city, and (c) deep discounts on Z-Wave/Vera merchandise for their own use and testing.

I love the online “virtual Vera” MCV now provides for system tests (though it could stand having more devices hooked up), as well as the “back door” component MCV can use for troubleshooting. How many other companies bother with such nice touches???

While we’re talking about things could improve… What about a much more exhaustive list of “Devices That Work With Vera” in the MCV Wiki?

We need to beg power users to contribute their knowledge to:
http://wiki.micasaverde.com/index.php/Supported_Hardware

That’s all good, but you guys answering question about easy maintenance - i.e. how to avoid re-doing the entire config if one has to replace faulty Vera with new box. I totally agree with guessed on the fact there’s no too many faulted routers (power supply yes, but not the router) stories floating around. They do, however, hang up or reboot without user’s consent, some do it quite often.

The original subject of this thread was not about easy fixing, but on how to make the overall system robust enough to leave running unmanned for weeks, with maximum self-recovery capabilities possible. I came on this subject while designing my VoIP system, which has similar problem - it depends on too many factors: power, connection, providers, carriers, etc.

325, I was specifically tackling/challenging the question about hardware reliability that you outlined here:

given inevitable problems with hardware, redundancy is a must, so there should be a kind of clustering option: redundant controller with identical scheduling, capable of verifying that the primary controller is alive and well, if not (router crashed, dongle locked up) it should be able to take over. Even better - the second controller would verify that scheduling works as expected.

the abridged version of my discussion is

“things break, be able to handle that quickly, but not instantly… since it’ll be way cheaper”

and a related “work on the software so they don’t break often”, which is something we’re all helping with here.

I happen to have a backup router, modem, etc, but I don’t believe that’s commonplace in most homes, and nor is any form of backup Network. Most people simply need to be able know that something is broken, and a way to replace it “quickly” when it goes astray.

I look at a home automation controller as something that automates activities that can also be done manually. When the automation controller “breaks” things can be done manually for a day or so until the replacement part comes along.

This of course, assumes it’s not breaking daily :wink:

- redundant internet connection, if primary is cable then there would be backup DSL or cellular...

There was an earlier discussion on getting a Cellular modem working. It was back in January sometime, but I don’t have the thread handy. The Cell-modem is expensive, and I don’t believe most will pay… not unless you have a McMansion, but then you’re probably buying a bigger HA system (Creston?)

A cheaper option would be to permit a local, direct-attached, Pay-per-use Cell-phone so that Vera can send Help messages when it detects something’s a miss. This would be generally useful for other things, and could be opened to also “receive” Emergency SMS commands (similar to the other thread) in cases where the DSL/Cable is down. It could also be used in Non-Emergency situations (“Alarm my house”) as long as there was some basic auth mechanism.

You’d still have service, just at degraded levels, until you address the Network connectivity problem… or the Burglar that’s cut your wires :wink:

Most of these low level interface components are available for generic OpenWRT (via gnokii.org). If these were re-enabled in Vera (hint, hint) then we’d be a step closer to a rudimentary backup to [local] Network/connectivity failure.

If one can do things manually (which means someone is physically present at site) he doesn’t really need that level of reliability. But this doesn’t cover the scenario where the person capable of troubleshooting of HA system is in 6-weeks business trip… overseas.
Having spare Vera in the closet won’t help much. SMS with “help” message won’t be very helpful either if internet connection at home went down and one has no other way to get in. I agree that any efficient solution will cost something, but I don’t think this entire subject is in “everybody” scope anyways…

My biggest concern with leaving Vera running unmanned for weeks is fire hazard. If for any reason it fails to turn off something it was supposed to turn off, it may be a problem: significant part of residential equipment isn’t designed to stay ‘on’ for too long, and doesn’t have adequate overheating protection.
What would be very useful if backup Vera box could monitor state of the system, if it sees massive failures, such as multiple schedule violations, or devices in wrong status, it could alarm the owner in any way, so he could ask someone to come, check things, and possibly shut everything down… better safe then sorry.

Theoretically this could be done in a luup plugin. It creates a timer that every 1 hour, for example, checks if device x is on, and if it’s on for more than 3 hours, then it sends a text message.