Goodbye Fibaro hello OpenHAB

[robmac 173]

 

Hi,

 

It is no secret that I have not been happy with HC2. It is long in the tooth but that is not the issue that finally made me make the decision to move.

 

It was the update last year that ripped my network to bits and prevented me putting it back together again that was the final straw. I have run a 140 node network for three years with very few issues but that was all destroyed in an update.

 

I started looking for something better.

 

Then a firmware that was pulled within an hour and that said it all.

 

This weekend I made my final decision.

 

I have decided to move my network to OpenHAB2. So why the decision to move?

 

• I can find no reason that the move will not be for the better
• It is a little more technical to run OpenHAB but the zwave implementation is sound.
• I can work with the team over there to make it the best z wave solution  in the world.
• I can write my own plugins....

 

What will I miss ..... only  LUA and some pretty pictures and the forum.

 

It will take me a while to move across with over 180 modules live in my house .  Despite having all those modules working together across two homes, I  have found that running side by side works fine. Certainly a lot better on OpenHAB but that is probably because my HC2 network is corrupted and too hard to fix. 

 

I will be posting how I get along on the OpenHAB forum still under robmac. I will be writing there how I moved from HC2 to OpenHAB. I will also add anything uefull that I can to OpenHab to make it easier for anyone wanting to follow,

 

This is not necessarily a permanent move and I will still be here to follow how things are getting on, keeping in touch and checking for module updates. 

 

If Fibaro were to come out with a 700 series solution that rocks I might reconsider. It would have to be the absolute best and somehow I can not see how they would achieve that.

 

 

Cheers,

 

Robert

Edited March 17, 2019 by robmac

[Piotr Bojko 27]

I am planning that move too. Only one thing (except time) scares me now - how to move my nodes to openhab. 

[robmac 173]

52 minutes ago, Piotr Bojko said:

I am planning that move too. Only one thing (except time) scares me now - how to move my nodes to openhab. 

 

See you over there.

 

My approach:

 

silicon z-stick upgraded to new firmware.

 

A few new modules. Set those up to form a backbone and started moving node by node.

 

My live z-stick gets the frankenstein treatment this week to remove some of the routing. All based on this... 

Please login or register to see this link.

 

My first attempt at this mod is not very pretty so I will not distress anyone with it.

 

 

[Piotr Bojko 27]

HC2 has an option to movev config to a slave. Just wondering whether this can be used to move a master from hc2 to a aeon stick (used by openhab).

 

 

[robmac 173]

25 minutes ago, Piotr Bojko said:

HC2 has an option to movev config to a slave. Just wondering whether this can be used to move a master from hc2 to a aeon stick (used by openhab).

 

 

Could try and see if it works. You would need to promote the aeon to master at some point and HC2 is all set for pull in only.

 

I can't see why it would not work but I am doing a slow migration bit by bit as it is better for me.

[AndrewB82 4]

How do you handle LUA scenes migration if any?

 

Best,

 

Andrzej

[Piotr Bojko 27]

The only option is to recreate them in openhab native solution I'm afraid.

[D-Vine 64]

Interesting solution that I already saw somebody else did on frensh forum. 

But for me the scary thing to do this or consider is some of the great LUA scenes people made here.

 

- check presence with phone detection (android)  

  I do this with UniFi MAC detection, with the combination of checking the motions sensors around the house

-  advanced light scenes (sancotronic very smart lights) 

-  telegram message 
  battery low, washing done, Plugwise day overview, SolarEdge Day overview etc.

- I have buienradar alert (dutch)

- Advance humidity check bathroom (on above xx value , and only for 90 min. for example)

- WakeUp light for my little girl's room with different time settings on different day's

- Automatic check Home/ Away and combine scene's with this settings?

- Foscam camera security

 

I see I can integrate

netatmo

Marantz

panasonic

SolarEdge

EVOHome

 

But can you read out all value's and make actions to this integrations?

or only like a possibility for stand alone read out and control ?

 

 

[1152 33]

If you have "sort of a working network", (which most of us have), based on HC2 and Lua scenes I believe the key to transition is as @robmac pointed out. You can actually operate two z-wave networks in your home. In principle it may not be the optimal solution. In a real world situation I would argue that it may be the best, at least for an interim transition period. This way you may take small and safe steps, avoiding a big jump, an if for any reason you decide to, be able to turn back.

 

19 hours ago, robmac said:

 

...If Fibaro were to come out with a 700 series solution that rocks I might reconsider. It would have to be the absolute best and somehow I can not see how they would achieve that...

 

 

Yes, I would be very tempted. Actually I may not be able to resist shiny new hardware pushed by some clever marketing. And then I would continue to include Fibaro fw/sw in my night prayer.

[dareszek 7]

As a FIBARO team will not deal with what works as well as possible, we all have to wait for it - it means to escape Sad, but true ...

@robmac, let us know what your transfer is and what is happening.

greetings

Edited March 18, 2019 by dareszek

[lasmartuser 1]

Why not hassio instead of openhab?

Edited March 18, 2019 by lasmartuser

[Piotr Bojko 27]

Hassio is built upon Home Assistant which I find more lively developed but far less stable. Openhab is considered more stable.

 

As for my personal opinion - I am Java engineer with over 10yrs experience and I am more familiar with Openhab tech stack (which is Java & OSGI).

 

As for LUA scenes for presence detection - Home Assistant and Openhab have this either native or via plug-ins which I consider more stable and still more native then Lua scene. 

As for LUA scenes in general - both Home Assistant & OpenHab have their equivalents. Even more - you can write your own custom plug-in for any need which is an ultimate power for extending your automation hub. LUA is albeit handy it is for sure not as powerful. HC2/HClite in general are not as powerful for extending. Which works for me in favor of OpenHab, because my engineering skills. One question remains - who is not when using Fibaro hubs...

As for telegram - both have such integration.

As for Foscam - OpenHab has such integration. But I personally use the Synology Surveilance Station which... has respecting integration in OpenHab.

As for reading from integrations like netatmo & more - all is possible because of the approaches in Home Assistant or OpenHab. They are either services or channels, just like any other things in HA or OH.

 

Another reason for me to leave is the new mobile app - OpenHab has something called sitemap which is convinient way to build you UI in a way you want. I would rather recreate UI paths from old Fibaro app then use the new one (which is slow and not as UXable as old one for me).

 

I have Chromecasts, Neato, Synology, etc - which... is either not available in Fibaro or works as a virtual device or scene which need to be maintained a lot. Those things are fluently handled as normal bindings/integration in OpenHab. And for sure - such integration already exists I don't need to beg Fibaro to create this or beg my wife to have a time to do it myself. 

 

 

 

[D-Vine 64]

21 minutes ago, Piotr Bojko said:

Another reason for me to leave is the new mobile app - OpenHab has something called sitemap which is convinient way to build you UI in a way you want. I would rather recreate UI paths from old Fibaro app then use the new one (which is slow and not as UXable as old one for me).

 

I have Chromecasts, Neato, Synology, etc - which... is either not available in Fibaro or works as a virtual device or scene which need to be maintained a lot. Those things are fluently handled as normal bindings/integration in OpenHab. And for sure - such integration already exists I don't need to beg Fibaro to create this or beg my wife to have a time to do it myself. 

 

Nice to hear and that you share your experience on this.

Question:
Recommended hardware. I know you can do this with a RasberryPi but I saw also it's more recommended to use something like a little a Gigabyte GB-BXBT-2807 or something.

And for me it's important for a external z-wave antenna because my hardware is in a steel 19"rack.
I only see USB z-wave antenna's for this?  

[petergebruers 1,261]

20 minutes ago, Piotr Bojko said:

Hassio is built upon Home Assistant which I find more lively developed but far less stable. Openhab is considered more stable.

I agree. The lead designer of hass expects "1.0" to land by the end of this year, or Q1 of next year, kind of admitting that things are in flux, improve and break, then the cycle repeats. Fixed release every 14 days, possible fixes in between (expect a bug to linger for 1-3 days)

 

I also like the summary you give of openhab, but each time you say "openhab", you could say "home assistant", because their capabilities are similar. This saves me a lot of typing.

 

I do think hass has more flexible design of ui, but as you say it is much worke in progress. If you want to know more, I recomrend to watch movies about hass "lovelace" inter&ace which btw has become standard now.

 

Based on tests of openhab and openzwave (in hass) I'd say Z-Wave is "best" in Fibaro HC, then openhab, then openzwave, but each can do something that the other cannot do. Based on code comments: first openhab, then openzwave. Based on "implementing the spec" and features, score is a bit mixed, but I'd give that title maybe to HC2 and Z-Way (don't know about homeseer, indigo, fhem or others), then openhab, then openzwave (which was based on reverse engineering and predates the release of the docs.)

 

My main system is HC2. But I run alternative software (small systems) "to put things in perspective".

 

You always have to be careful not to jump to conclusions, based on only 10 devices though. Any controller can handle 10 devices.

 

If you monitor the forums of those products, you'll see what I mean.

 

13 minutes ago, D-Vine said:

Recommended hardware. I know you can do this with a RasberryPi but I saw also it's more recommended to use something like a little a Gigabyte GB-BXBT-2807 or something

For testing, Pi (possibly a clone I& you are linux adept) because nothing beats price performance.

 

For production: nice low power intel board with SSD! Like the HC2 (but with newer internals)!

 

But ARM is catching up, next year things might be different.

 

16 minutes ago, D-Vine said:

And for me it's important for a external z-wave antenna because my hardware is in a steel 19"rack.

People who know me will surely confirm that I say "RF is vodoo"

 

So try the standard antenna first!

 

Robmac, like many others, has asked me if he can improve things with "a better antenna".

 

The short answer: probably not, because "high gain antennas" also make it more directional. Adding extra cable also quickly dissipates whatever gain you have...

 

RF is vodoo. This guy tested some 868 MHz antennae:

 

 

 

The performance of a Z-Wave dongles is not ony about the "antenna part":

 

If this topic interests you, I used a "dongle" to map the range of two devices in my house:

 

Click on the spoiler in that post to see the maps.

[Piotr Bojko 27]

Thx @petergebruers for your feedback.  For sure HomeAssistant and OpenHab have similar capabilities, I am just better in Java  . 

 

I want to move about 80 z-wave nodes installation from fibaro HC2 to OpenHab.  Having your thoughts about z-wave nodes performing better in HC2 than openhab makes me think about possible issues I would encounter in my plans. Could you elaborate more on this? Thx in advance. With Fibaro I have some timeouts with locks. I hoped that I would encounter issues with no higher severity in OpenHab.

 

As for my hub setup - I was planning to use NUC or something similar. 4 cores - because is is Java and I would go with a DB and some side software like grafana. Small SSD. Aetoec Z-Stick. Little more power to handle more things in price slighlty 2x more than PI.

[D-Vine 64]

1 hour ago, petergebruers said:

 

My main system is HC2. But I run alternative software (small systems) "to put things in perspective".

 

So try the standard antenna first!

 

The performance of a Z-Wave dongles is not ony about the "antenna part":

 

Thanks for sharing!

For me I just consider what a other system can do for me, Is it better, or better for the future because of the amount of systems that can be integrate?
I like fibaro a lot, but looking at other systems like openHab I think why why does Fibaro not have those amount of standard integrations for system. 

I my opinion the plugin section should be the one for it, but I don't see the development that takes the HC2 to that level. Oh this is nice, this is handy easy to use those plugins let go. 

We trust on the top programmer like you, sankotronic as example and many more for building great scene's but also to integrated other systems, but I miss the part of Fibaro here.  (for example sankotronic made Netatmo VD better then the default plugin of fibaro)

But at some point I have a feeling Fibaro is up to something making the system better with other GUI (like the new app) and more. And as always I need to be patient     Sometimes it is itching to switch on other system

[petergebruers 1,261]

5 hours ago, Piotr Bojko said:

I am just better in Java  . 

I understand, but everyone has to decide for himself if the language of the underlying code matters to him personally. Maybe it does, maybe it does not, let me give a perspective...

 

The Tiobe index tells me:

 

Please login or register to see this link.

 

Java, C, Python, C++ and VB.NET are the top big languages (caveat: based on their definition of the success of a language) so they have possibly a large number of contributors.

 

If you know Java, you'll be comfortable with C/C++/C# and less with Python. But since you use Lua... you'll understand Python.

 

If I ever write my own system (which is not going to happen) I'd use Haskell (43 on the list) or Erlang (49 on that list) because those languages are totally unlike the top languages, so I do not have to fear comments from other coders about my bad coding style. LOL.

 

The first level: do you really want to understand the code? That's a learning experience, and I recommend that. It would not matter much if it were Python or Java, those HA systems are big, no, huge, and navigating the code and forming a mental picture of the architecture is a real challenge.

 

The next level: do you want to write code? In that case, learning a new language might be a hurdle. And learning the debugger, stepping through the code (it is "async" and often "multithreaded" so stepping will do unexpected things. You'll probably better of tracing stuff).

 

The ultimate level: contribute to the development of the product or a plugin (I am not talking about "a motion control scene" but support for "your new heating system which has no support yet"). Or you want to fix a bug, are you ready to put it in a repo fork, learn and fight git, follow the coding lines, discuss that with the maintainers, finally get your pull request merged then find out that your new code breaks things for some users? It is not only writing code + also support users and maintain. Don't count on a large number of people helping you out, go check the number of maintainers in the code. In fact, you might be lucky if you find one.

 

5 hours ago, Piotr Bojko said:

Having your thoughts about z-wave nodes performing better in HC2 than openhab makes me think about possible issues I would encounter in my plans. Could you elaborate more on this?

 

I would not use the words "perform better". Part of the problem with large networks is the number of "interactions" increases and timing gets more critical, you get more timing issues and it has to do with reporting settings of devices and how you write your scenes as well. Z-Wave is not a fast network.

 

I can think of an example where different gateways work quite different though, but it is a rather specific example but uncovers different philosophy behind gateway. It is about how they deal with a Z-Uno

• Z-Uno is a special device because it ALWAYS tells, by IDS "I am a Z-Uno". The IDS of these devices are fixed. But a Z-Uno can represent almost any kind of device or combination, eg a relay and a dimer and a temp sensor. This should not be a problem because Z-Wave reports the channels, capabilities, ranges, and a bunch of other stuff.
• Z-Way is from the same company that designed the Z-Uno so no surprise their gateway supports it. But the underlying reason is interesting: they have the concept of "interview" which means the gateway asks about the capabilities. So you can have any number of Z-Unos and they can all have different relays/dimmers/sensors
• AFAIK HC2 uses the same method although you do not have a "interview" button. I think the equivalent is "(soft) reconfigure" and HC2 does that at inclusion time. So HC2 too can support multiple Z-Uno, each having different setup
• OpenZWave uses the concept of "template" combined with a "settings database". So initially, it applies a template to the device and this gives you the option to change parameter 2 "Blink Service LED". Good. 
• Openhab does not interview devices, it uses a template (database entry). So it can only support one Z-Uno with a specific set of command classes. In fact, it is almost correct to say that openhab does not support a device, if you cannot make a template which can be used by all users. I checked, on feb 19 of 2019 the maintainer confirmed and posted 

  Please login or register to see this link.

   - I quote "For the Uno, as far as I understand these product IDs can not be changed by the user, so ALL devices will look the same to the binding."

Please login or register to see this link.

 

I cannot tell if openhab with 80 nodes is good. I have HC2 with 90 nodes( was 100 a few months ago but I moved some to test) and it is good (which surprises some people, but it is true..).

 

5 hours ago, Piotr Bojko said:

As for my hub setup - I was planning to use NUC or something similar. 4 cores - because is is Java and I would go with a DB and some side software like grafana. Small SSD. Aetoec Z-Stick. Little more power to handle more things in price slighlty 2x more than PI.

FWIW I think Z-Wave.Me updates their controller firmware more often than Aeotec but I am not sure if that is important. Check the latest FW and check SDK of Aeotec.

 

The issue with PI is usually corrupted SD card and SD performance because they are not meant to be used in that way and quite a few posts on those opensource fora deal with "I have corrupt files" which is either SD card, power supply or overheating or a combination of that.

 

There are other ARM-INTEL SBC differences as well, I can go on for hours about systems.

 

Yes, influxdb  + grafana!

 

4 hours ago, D-Vine said:

For me I just consider what a other system can do for me, Is it better, or better for the future because of the amount of systems that can be integrate?

I think a HA system is "insanely complex". It starts as a remote control for a dimmer and a plug... It ends with Sonos integration, Temperature Control, Alarm system(s), ... and a ton of rules.

 

I really, really, really cannot summarize HC2 or OpenHAB or Home Assistant in a few sentences.

 

As you say: "the amount of systems that it can integrate" certainly is a big part of it. They do not even see Z-Wave as a privileged platform, it is "one of the possible integrations".

 

I do like the way Z-Wave gives consistency. You take it for granted that a battery level is 0-100%. But other platforms report "Volts" or even "good" or "low".

 

4 hours ago, D-Vine said:

(for example sankotronic made Netatmo VD better then the default plugin of fibaro)

I don't have Netamo but I trust you are right, and I do not know what happened to the original "plugin" concept. It started out as "open" then was no longer available to users, in fact it disappeared before I could build my first plugin . I am unsure why his code was not promoted to the "official" release, it is an interesting question. Tagging @Sankotronic 

 

4 hours ago, D-Vine said:

Sometimes it is itching to switch on other system

Maybe not "switch" but try something else, "to put things in perspective" as I always say. You might still like HC2 and decide to stay. I wanted to toy with Ikea Tradfri and I run HASS to control one lamp... LOL. There's a lot out there, begging for your attention...

[robmac 173]

7 hours ago, petergebruers said:

I am unsure why his code was not promoted to the "official" release

The plugins execute at a high privilege level so a plugin could be used to compromise the security of the system.

 

7 hours ago, petergebruers said:

I cannot tell if openhab with 80 nodes is good. I have HC2 with 90 nodes( was 100 a few months ago but I moved some to test) and it is good (which surprises some people, but it is true..).

I think most people admit that it is zwave that is difficult with over 100 nodes. You need luck and voodoo or tooling. HC2 just lacks the tooling and it appears I lack the voodoo on this version..

 

I had luck and voodoo that got me to 140 nodes with good reliable performance but I have lost the voodoo or luck and I can not get it back. I fix one issue and something breaks at the other end.

 

7 hours ago, petergebruers said:

Openhab does not interview devices, it uses a template (database entry). So it can only support one Z-Uno with a specific set of command classes. In fact, it is almost correct to say that openhab does not support a device, if you cannot make a template which can be used by all users. I checked, on feb 19 of 2019 the maintainer confirmed and posted 

Please login or register to see this link.

 - I quote "For the Uno, as far as I understand these product IDs can not be changed by the user, so ALL devices will look the same to the binding."

Having looked at the code it could be changed.

 

There is no haste needed. You can run both side by side. I am sure when I get HC2 down to only my old nodes that I do not want to upgrade yet it will be fine.

 

[robmac 173]

I thought I should clarify a bit more regarding going beyond 100. All of this is theory as HC2 does not have the level of logging to help understand.

 

1. How did I get 140 nodes stable in the first place.

• I tuned the traffic down as I went
• I made sure the network had lots of time to learn as it grew. Increased the number of polls and reports in areas that had issues and reduced in other areas so the routes were learned. 
• Only reconfigured a node on the mesh when nothing else was busy, only if it had issues and only one at a time leaving time between and lots of time not 10 minutes, I left hours. I believe this allowed good lists of neighbours in the nodes so the neighbours that form the backup route were all well set. So when the last working route failed, the one it tried next was likely to be OK.
• My theory is that it is the dynamic routing that causes issues as the number of nodes increases and the traffic saturates. A node starts to fail to manage to use the same route more than once but some of the failures are congestion not  issues with the route. I am trying to get my head around if this is true but when you look at a zniffer file and what you get as info from Fibaro not a chance.
• Of course if it fails to use all the configured routes you eventually get explorer frames and these appear to be hoggers of bandwidth. This in turn makes more nodes fail on their normal route etc etc.
• As things got more extreme (155 nodes at one point) I asked how often I needed an update from some nodes. If it was one time a day then I removed the lifeline and polled one a day.
• For some Fibaro movement sensors on the edge of the network. I used direct association to the module they triggered as it was a direct link not routed. I then  removed the lifeline from the sensor. The controller does not know the alarm was triggered but it knows the light is on. Why an advantage? Only one set of reports across a couple of hop route. Every little.
• and a few other tricks
• eventually tuning down the polling so very little and none to some nodes.
• tuning down the energy and temperature reports to the minimum I felt was useful
• I think I could even have have added more node back. The 15 removed were when a bit of the house was being renovated.

and it worked really well for three years.

 

So what went wrong?

 

Well I think two big things but possibly more. Possibly something else and these reason are just coincidence but puppies and piles of goo comes to mind:

 

1) FLIRS not sure how but am investigating 

2) Full mesh rebuild that my network was too busy to support at 2am every day that caused my HC2 to lock for the first time since I can remember. I think this made a mess of a lot of those neighbour lists and hells knows what it did to the routing and possibly other corruption .

 

So why do I not do it again?

• TRUST: It is gone and the work to fix is huge. If I invest in this again and they release another firmware that does stupid?
• My wife would probably murder me. She loves this when it works but it is all my fault when it does not.
• I don't think I could put the effort in again. It was like making a beautiful painting but then somebody walks through the door and throws emulsion paint all over the canvas. You can clean up the mess but do you have the heart?
• It is just too difficult and not the adventure it was first time. Some things are fun once the second time they are a @$£%^^.

Possibly this pull of the latest firmware is a good thing and they are finally taking quality of releases seriously .

 

Will it be better on openHAB?

 

In reality the same things limit zwave but yes it will work for me because it is an adventure again and I am in more control. Nobody restricting me and implying I am not big enough to mess with this or that. No limits on what I can do with my code and platform. 

 

Will it work? What about the bits Peter says are missing?

 

I am in control I can add all the bits that Peter points out are missing in Zwave if I indeed need them. Don't be confused, there are lots of people with 120+ nodes on openHAB and if you look in the database it often has templates before HC2. All of the ones you have in your network are probably there and a few that HC2 does not.

 

openHab Zwave does lacks complete handling of endpoints and is also locked so if there is no template then you can not add. Actually neither difficult to fix.

The lack of a template is easily fixed with no coding as you can request to the team and they add fast but also you can edit the files in the JAR so that you can use immediately.

 

Is it for everyone?

 

No but if you want to do more than the basics or have a wide range of integrations you should look at it. You do not need to understand the code. I am sure most people that use it do not.

 

What do I think about HC2?

 

It was great and I enjoyed it. 

   The high point was user plugins and the planned market for them so people could extend to all of those new things.

   The low was that release that smashed my setup to pieces and stopped me uploading my plugins.  Yes I had plugins.

         I could root my HC2 but that puts me in a horrible place. Anything I do gets overwritten if I want a Fibaro update. What is the point if I have some but not enough control?

 

HC2 for me was a missed opportunity.  With all of the advanced features nobbled I feel it is trying to be a solution for the mass market that is plug and play but 100+ nodes plug and play on zwave? Sorry not going to happen any time soon.

 

Doing all of the third party integrations inhouse? Just how big is the development budget would you need?  I think how many of them do not work is testament to the folly of this. Most people have had to resort to virtual devices.

 

UNbless they change direction my gut feel is that HCX will probably be like smartthings but without the mass market marketing budget or huge development budget.  Possibly Fibaro should back out gracefully and stick to making great modules. Make the firmware for these OTA on all other controllers so they can sell lots of them and get out of the controller game altogether.

 

As always a bit of a stir up.

 

Expect to be here for a year or so more as I am in no hurry to move all things across and I still need my HC2 for firmware updates. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

[robmac 173]

And this is how fast template update should happen thought you can add to the site yourself.

 

Please login or register to see this link.