Odd behaviour when power is lost for Z-Wave devices

[raym]

Hi,

I've been doing some resilience testing on my z-wave network I've discovered what I believe is a fairly significant issue related to when power is lost. As an example, I tested cutting the power to one of my mains powered plugin switches. The sequence of events which followed were:

1. DZ soon detects that the device is offline: OpenZWave: Received timeout notification from HomeID: xxxxxxxx, NodeID: 2 (0x02) - GOOD
2. From this point, toggling the state of the switch appears successful in the UI, despite there being no successful ACK received from the device - BAD
3. DZ eventually marks the node as "dead" and changes the device widget in the UI to "red": OpenZWave: Received Node Dead notification from HomeID: xxxxxxxx, NodeID: 2 (0x02) - GOOD
4. DZ never restores the state of the switch to alive (green) after power is restored. A restart of DZ is required to force node re-discovery - BAD

Point 2 is an inconvenience but not a major issue as the node is eventually marked dead when a timeout is received. Point 4 is the real problem in my opinion as the device remains unusable until DZ is restarted. I should point out that this device has polling enabled too so I would have expected DZ to eventually discover when the node comes alive again automatically. In any case, it seems unusual that a restart is required for this to occur.

Happy to provide more info if required but it would be nice if some slightly better fault tolerance could be built in to cover scenarios such as power loss.

Many thanks.

[Egregius]

True
Also 'Heal Node' in case that it's marked dead doesn't work to bring it back alive, only a reboot does.
I've been struggling lately with marked as dead nodes. They apear randomly and happens on many different main powered devices.
Because of that I decided to go ahead with a clean install of Domoticz and a hard reset of the controller. Unfortunatily it didn't solve it.
In one of the + 3*** versions there was a big update of openzwave, I guess the issues started with that.

[jannl]

I have a greenwave plug, with polling enabled the node comes back to life again.

[raym]

Even after its been dead? For me, if DZ marks a device dead, then that's it... Until I restart.

[Egregius]

Are you on beta 3867?
I returned to stable 3530 and that one has another openzwave version.
Will see if the dead nodes return...

[jvdz]

Even after its been dead? For me, if DZ marks a device dead, then that's it... Until I restart.

Did you also manually switch the ZWAVE unit On/Off to send a trigger to DZ of it's presence again?

Jos

[Egregius]

Aarrgghh, the Greenwave 6 was just marked as dead.
A press on the button on the sides makes it back alive. But that's no good workaround? It should be possible to reactivate it automatically without user action.

[jannl]

You do have polling enabled for the greenwave plug?

[Egregius]

No, no polling. The Greenwave has worked fine for weeks without polling.
Marked as dead happens randomly on Greenwave, Fibaro Wall plugs and Qubino dimmers/relays. As it happens on different makes/models I don't think I can blame it on them. Has to be something within openzwave/domoticz.

[jannl]

Was solved by me by enabling polling

[Egregius]

OK, enabled it for now on 3 devices. Had 3 marked as dead after last reboot
If polling solves (workaround) it I believe that polling should be automatically enabled when a device is marked as dead.

[Egregius]

More than 10 minutes after enabling polling: no result. They are still marked as dead.
Heal Node and Heal Network doesn't help either.

[raym]

More than 10 minutes after enabling polling: no result. They are still marked as dead.
Heal Node and Heal Network doesn't help either.

Yeah this is the point. Even with polling, it makes no difference for me either. A restart of DZ or system reboot fixes it.

Running the latest beta: 2.4020

[raym]

Even after its been dead? For me, if DZ marks a device dead, then that's it... Until I restart.

Did you also manually switch the ZWAVE unit On/Off to send a trigger to DZ of it's presence again?

Jos

Ok, so this works! Thanks.

Don't you think though, if polling is enabled, DZ should detect the node is alive again automatically?

[raym]

Let me be more direct ....
Would it be possible please, to restore a dead node automatically whenever a poll request is successfully acknowledged? This would alleviate the need for manual intervention when a node goes offline temporarily (such as during a power outage).

Thank you in advance.

[zaadstra]

Hi,

I have the same issues, and wrote about that earlier. To me, the Greenwave appears as an unreliable device and on top of that, OpenZwave does not handle it properly. I totally agree that if a node is marked dead, it should be tested once in a while if it is back in the land of the living.

The discussion on the OpenZwave forum ended like this: our (OpenZwave) software is perfect, so we ain't gonna fix it. I do not agree with that.

Here, the marking dead occurs totally random (so it seems to me), sometimes many weeks no issues and then more than once on a day.

I discovered that pushing the small button at the side (near the on/off button and dial) also wakes up the connection to Domoticz. It takes some seconds and then the green ring blinks.
I have also seen that enabling/disabling the OpenZwave log (option 2 in the Z-Stick S2 parameters) makes the Greenwave alive. In the OZW_Log.txt I see that the OpenZwave stack is restarted ( tail -f OZW_Log.txt ). This proves that not the Greenwave has gone nuts but the OpenZwave stack pessimistically marked it as dead because of 3 comms failures.

In this other topic a script is presented by user chadweek that checks on the Greenwaves and restarts Domoticz. This is just a plaster but at least the Greenwave will be controllable most of the time.
I've never used these pieces of junk (unfortunately I've got two of them) because I can't rely on them, you never know if they miss scheduled switch times...

viewtopic.php?f=24&t=7024

[Egregius]

To me, it only appears after a restart of openzwave.
Sometimes everything works, sometimes 1 marked as dead, sometimes 3-4-5.
Until now another restart was the only solution.
Just tried the debug thing and it wasn't marked as dead anymore.
So, will try to implemented that in a script and see what it does upon the next dead marked node.

[zaadstra]

By the way, the Greenwave is capable of more misery ...

Socket 1 is not responding to on/off commands. It does report metering and it does switch on/off with the button on the device. The actual on/off status is not reported however, like socket 2-6.

Issuing a switch command notes this in the Domoticz log:
2016-01-11 20:26:38.685 OpenZWave: Domoticz has send a Switch command!, Level: 0, NodeID: 9 (0x09)
2016-01-11 20:26:48.716 OpenZWave: Received timeout notification from HomeID: 25480967, NodeID: 9 (0x09)

And in the OZW log:
2016-01-11 20:26:48.714 Error, Node009, ERROR: Dropping command, expected response not received after 1 attempt(s)

Only once (?), and it does not declare the Greenwave dead.

Oh my, after some reading in other Greenwave posts I got the idea to switch in the OpenZwave Control Panel, and there the switch 1 properly responds to the command. So this is a Domoticz thingy?

[Egregius]

Oh, enabling/disabling debug restart openzwave and all nodes are queries again. That explains why it sometimes works after that.
First attempt of the script crashed domoticz, so I'll keep with the reboot.

I will try ozwcp next time it's marked as dead. Does it came alive then?

On the other hand, it doesn't only happen on the Greenwave 6 plug, also on Qubino relays, Qubino Dimmers and Fibaro wall plugs.

[zaadstra]

I will try ozwcp next time it's marked as dead. Does it came alive then?

I don't know really, did not try that yet. I see the Control Panel is a powerful screen, capable of breaking many things

At the moment I'm only struggling with socket 1, very annoying ...