I’m running into an annoying issue with delays between zwave devices. Delays are easily 8 to 12 seconds by now.
My system layout:
• Running on Synology RS1221+
• HA in Docker
• Z-wave JS in docker
• 34 Zwave devices
• Server is in the garage (maybe the outer wall is too thick?)
Symptoms:
• Motion sensor to light delay. You stand like an idiot in the dark waiting and counting for the light to turn on.
• Other scenario. You press the Fibaro Dimmer switch. Light returns to previous brightness (factory default). After 8 seconds the light dims to the value you prefer when you press the switch. Another few seconds pass by and the light toggles off. Again, you stand like an idiot in the middle of the dark room. Walk back to the switch in the dark and press it again. This time you keep standing next to the switch because you’ve just entered the game “how many more times shall I be in this man-vs-machine endurance race”. In other words, this scenario loops a few times before you end up with a well lid room.
• My wife is really starting to dislike our dumb-smart-home up to the point I’ve had to kill certain automations. (If I’m lucky it stays with these assassinated automations, I hope I can revive them after performance is better. At one point we had a toilet disco)
I’ve rebuild the routes in zwave js.
Thoughts:
• Would moving the wave dongle to inside the house solve it? (Are there zwave ethernet gateways?)
• Would 34 zwave devices be too much for one dongle?
• Is the Synology NAS not powerful enough?
• Is the zwave network constipated, too crowded?
• Are there devices spamming the zwave network? (I’ve had once a zwave smoke sensor drain it’s battery in two months)
• What would be the bottleneck, how to check and how to solve
• Oh, could other items in HA also slow things down?
First of all. Are de devices battery powered? Do you have devices that act as repeater or are the devices all connected directly to the controller? Are the devices added securely with S0 or S2 or without security?
What kind of controller stick do you use? What kind of devices ( and what brands) do you have added to the network?
You could check the logs of ZWaveJS to see of there is a lot of traffic going to the stick. It’s not ideal, but removing all devices could be an option. And then start with 1 device. Check how it responds. Then add another, etc. etc. If after a certain device the other included devices stop working or are getting slow, you possible have the problem found.
You could also try to install ZWaveJS on a separate pi to see if that works. Perhaps that will fix it. Then that would mean the NAS is too slow. But you could always check the CPU and memory usage of the NAS.
34 devices should not be a problem. I have 64 devices added to 1 controller, and it’s still responsive. No issues with that.
Look at device stats and check for retransmits and failed tx/rx
The devices showing high failure and repeat you can try individual heal and see if that helps. If not next step you can try add range extender(I prefer aeotech) or add AC powered devices.
If devices are over 2 hops from controller you can try to reduce that but that’s just guess and subjective
-Not there yet in your troubleshooting journey.
-I run ~100?
-Possibly but incredibly doubtful
-constipated… good word actually <<< likely first step because you also said::
-zwave smoke sensor drain it’s battery …they do that when they go spamalot…
-do what @tmjpugh recommended and
-technically yes but check these first before looking at HA WAY more likely. Dying device spam, chatty zwave power meter (device in general), and s0 when not necessary all are known contributors for spam your network to death,
I have personally experienced all three at least once over the last 7 years.
Only a few are battery powered. 5 tops? Rest of them are mains powered . A few of them are both battery and usb powered. But these are currently powered by the usb port, without batteries in them. (listed below)
The controller is a Aeotec Z-Stick 7. The main devices (17 pc) are Fibaro Dimmers and Double Switches. Then come the AEON Labs MultiSensor 6 and a few Fibaro Smoke detetors. Rest are some remotes.
Average CPU is 6%. 16% of the 64Gb RAM is used by the NAS. (Not just Docker image)
Only a few timeouts and failed tx/rx. Maybe one or two every few devices. Secretly hoped to find one with a large amount, I think. Also expected to find higher numbers with the bad response times.
Eyeballing the graph, the max amount of hops is two. But I did not know it also uses the now usb powered devices. They are primarily used as battery powered. (the multi sensors)
No, I think no zwave device in my network is.
I’ve checked the events with a wildcard in HA. More quite than I would have expected. I did disable the Unifi intergration. I don’t actively use it and this did gave a lot of events of the dreammachine and network device trackers. Might not be related, but I did not use it anyway recently.
What would be the best way to monitor zwave traffic. I currently look at the events in HA. But I assume zwave JS also has a way to monitor. ( I expected the debug tab. But that one stays empty after firing a few zwave buttons)
The name in zwavejsui becomes device name in HA. This happens
—when device is first added to zwavejsui(you are able to name device when adding)
—when integration is first added
I have learned that you can basically name all devices as they should be in HA then remove/readd integration and HA will set names. If I rename an existing device I remove/readd integration.
A light in zwavejsui named kitchen becomes light.kitchen in HA
A lock named front door or frontdoor becomes lock.front_door or lock.frontdoor.
Are the delays more or less correlated to the right hand side of your graph?
Or specific to a branch of device type / vendor of the device…?
If they are random in nature… here is what I would do next:
Go into each individual devices, and change the parameters around the reporting frequencies, while also considering the use case of each.
For example, you can ask yourself: do I really need to know temperature and humidity changes by the minute? Can I change the frequency to every 1200 or 3600 seconds, or maybe for those light switched can I change the frequencies to 43200 seconds? Do I need power/energy consumption sensers telling me the updates all the time? Do I even need to know how many watts the 3 LED bulbs in the guest room by the minute? Would reporting every 10 minutes still be fine?
Note that many of those reporting interval settings are for “reporting back when nothing happens”. So lux sensors in dark rooms do not need to report back if nothing changes. Typically if the manufacturers design their products well, they will still report back immediately when something changes. You can run some tests.
If there are 2 temperature sensors in the same room within 5 meters, could I turn off one of those or do I need both?
The point of the exercise is really to think hard and see what exactly you need those devices for, and based on your intended purposes for those, you configure them accordingly.
===
If I’m in your shoes, I will turn off all the power / energy monitoring sensors for all your Fibaro switches, and turn on only if I absolutely have to… at least for a week or two, and see how it goes.
