Dashboards killing zWave

Do you have more on a websocket issue??

I've used WebSocket Client to watch each of my Hubs and although the App isn't all that helpful, it adequately displays the stream.

I use Websocket between all 8 of my hubs, 4 Hubitat to the NodeJS server and Homebridge, and a pair of Hubitat hubs using WS directly without any losses I'm able to detect. Of course I have Dashboard too, although it's active on a browser only a few seconds a day. That's a total of 12 Websocket connections, 11 of them in use 24x7 -- which makes me really, really interested in any Websocket issues :smiley:

Do you have a external stick??
Its the combination of external stick and dashboards

Can't say I know why this slowdown happens yet, but I've created a GitHub issue so it gets tracked.

I have a C-3 and a pair of C-4's... so yes, I have external sticks. :slight_smile:

However, the one running Dashboard has the stick(s) disabled.

Hey Team...Just catching up from my earlier post. Crazy times here in California.

Anyway, I did want to share my Conclusion, Speculation on the issue, my setup, and testing

Conclusion: Several simultaneous Dashboard connections (Client to Hub or Client to cloud) KILL ZWAVE DEVICE Connection. The Hubata application on the c7 loses the ability to effectively send commands to the end zwave devices (see testing note below).

Speculations: Based upon WireSharK Data, Websocket in not configured correctly or has a bug (agree with community members above - web socket issue) I believe the web socket issues disrupts/overloads the internal connection between the Hubitat application on the c7 device and the zwave radio on the c7 device. I am not a web socket nor hubitat internals expert but i see no need for 50 TCP transmission every 10 seconds for the TCP stream keeping the dashboard updated (this was for a 3 tile dashboard) - aka the dashboard socket connection is extremely "chatty"... perhaps someone left debugging on or set the keep-alive extremely short (ms ver sec perimeter) - again, not an expert.

Thanks for listening... See details below

Paul

Setup:

  • Zwave Devices: 64 (50 zooz switches v1/v2, 5 zooz 4-in-1 sensors, 5 v3 August Locks, 3 Farbro Leak Sensors, and 1 WaterCop Valve)
  • Zwave health: All optimal routes, perfect stability, no phantom devices (thank you for code version 142 and 145)
  • 4 "Panels" in 4 rooms: 4 FireOS tablets, running fully brower each pointed to a unique Hubitat Dashboard
  • Hubitat Dashboards: 26 Total with every Dashboard only subscribed to the devices needed for their associated tiles (as recommended by many of you above). The 4 "Panel" Dashboards are disabled prior to testing.

Testing:

  • All 4 room "panels" disabled
  • Add first room panel (enable dashboard in hubitat App)
  • wait 5 minutes
  • Zwave devices slow but work
  • Add second room panel (enable dashboard in hubitat App)*
  • wait 5 minutes*
  • Many Zwave devices fail*
  • Add 3rd room panel (enable dashboard in hubitat App)*
  • wait 5 minutes*
  • All Zwave devices fail*
  • At this point disabling the dashboards associated with these "panels" may not work and a reboot may be required*
  • Significant note: During the event, physically changing the zwave device always showed the proper position of the switch/device in the Hubitat App. The issue is Hubata application on the c7 loses the ability to effectively send commands to the end zwave devices.*
2 Likes

Thanks, Paul.

(1) It’s suspicious that the number of TCP transmissions is the same as the number of Zooz devices. Is that an exact number match?

(2) your notes say that the transmissions are client to hub or client to cloud. Are the endpoints always the same? I assume, when you say “client”, you mean the FireOS tablets? I assume they are connected to one of the two URLs (WAN or LAN) for the appropriate child Dashboard app, correct?

Really thorough testing. Again, thanks.

See below:

(1) It’s suspicious that the number of TCP transmissions is the same as the number of Zooz devices. Is that an exact number match?

  • It is not a match. Same thing occurs for any dashboard. I was only making the point that it's a lot of chatter for such a small dashboard (3 tiles). Interesting fact; the dash board running on my "panel" doesn't need to have any zwave devices... it still crashes my Z-Wave "system"

(2) your notes say that the transmissions are client to hub or client to cloud. Are the endpoints always the same?

  • Doesn't matter where I connect my fire devices (cloud or local) still crashes zwave

2a) I assume, when you say “client”, you mean the FireOS tablets? I assume they are connected to one of the two URLs (WAN or LAN) for the appropriate child Dashboard app, correct?
-Yes, my clients are the "clients" to at the associated dashboard, meaning they have each opened a web socket to my hubitat (local or cloud). Sorry, network guy - Client server...

One More note: I loaded the HousePanel app and set up the Node server on a raspberry PI. This system works a lot like actionTiles where a stream keeps a central server updated with everything on the hub and then only pushes changes to the end clients (aka panel). I tested this and never had an issue with zwave which makes me further believe this is a web socket issue... again, not a expert

1 Like

Make sure you don't have "Use all your devices" checked on your dashboard.

These settings on the dashboard could also impact the amount of data.

image

Thanks for the tip - I was aware of this and made sure that was the case when I did my testing. See my "setup:" for my test (above)... I can also confirm I do not have that checked for any of my dashboards.

I played with that too.. change local to 10 seconds... still crashed. More testing to do in this area but I thought I would share my results as I think their is some type of bug/issue.

I'll try running a bunch of dashboards and see if I notice anything. I'm currently on .145. Should be easy to test here.

Stable so far...

I am on .145 too. Hmmm... good experiment. I have not tried this experiment from a laptop... meaning running all four "panels" from my laptop. Give me a few and I will check.

No noticeable delay on virtual devices... let me setup a zwave test.

image

Still nothing weird happening here. I have dashboards open on 4 phones also.

Maybe a driver issue with one of your devices and it either sending or getting an unhandled response or something on a device refresh. For comparison, I have about double the devices that you listed. Ok, no more testing wife just came in and asked why the hell I'm turning on and off all the lights in the house over and over.

1 Like

@lewis.heidrick you're a genius... My "Speculation" about this being a hubitat issue was incorrect. This must be related to the fireOS table and/or Fully Browser (so many settings in Fully). I have eight dashboards open on my Laptop and my z-wave is fine. More work to do but I'm getting closer. Thank you!!!

1 Like

Np, hope you can track it down!

@lewis.heidrick You are a pro and thanks for suggesting a new path. Issue seems to be an old version of Fully Kiosk Browser. I was running 1.24 which I believe was not fireOS specific (loaded 2+ years ago). Once I upgraded to 1.40.3 for FIreOS it's been stable. Not sure how/why this would affect z-wave but once again, not an expert at either.

@mark.cockcroft... Can you check your FK version.. see if that fixes your issue

@danabw one more Panel joke before I put solved on this one? haha

Thanks again for everyone's help. My system is stable and my transition from SmartThings is complete! Peace has resumed in my house... ommmm

3 Likes

Awesome! I know how frustrating it can be to get a new toy and it just won't work the way you expect. Then buyers remorse kicks in. I bought a new DJI Mav 2 drone and for the life of me I couldn't get it to fly right. Tried everything I could think of. Turned out to be a buggy firmware. After that it basically flew its self. It took me a while to get everything dialed in and now it's to the point it's almost too fast. I have a rule that turns off all the lights and arms the alarm when me and my wife both get into bed. I rolled over onto my wife's side to grab the baby monitor and it triggered the rule and shut everything in a few milliseconds. My wife was in the living room at the time and she said "The power just went out in the house".

3 Likes