I am currently thinking about going Virtual Machine (on Proxmox) with my HA setup. I love the ease of migrating VMs from one proxmox host to another in the cluster, with some cleverness, one might even hav some failover. But there is one thing that ties me down to physical machines: The USB attached Zigbee-Dongle. If my Hass VM would migrate to another host, it would lose the Zigbee USB dongle. Yes, I could probably go the LAN-connected Zigbee gateway route, but the USB-dongles with ZHA have been so troublefree for me, I am thinking about something else:
I have read that you basically can âcloneâ Zigbee dongles so that to the Zigbee network they look identical, see e.g.
Would it be possible to take e.g. 2 USB sticks like the SkyConnects or the Sonoff ones, configure them to my current EUI64/extendedPanId, so that they look identical, and connect one of these to the active Home Assistant via USB passthrough, while the other one is inactively waiting on another Proxmox cluster machine without Home Assistant, waiting for its turn to be activated if and only if the HA VM is migrated to that host?
Could that be feasible, is it even worth trying? Of course it is more or less wanting to play with this kind of failover setup, less practical necessity (plugging the stick to another machine is not that hard), but it would interest me if this might work.
One thing I am not sure of is: If I âcloneâ the Zigbee stick (copy EUI64 and extendedPanId) are there other characteristics the two USB sticks will differ in, so that the VM will not be able to take over with the new dongle, like USB serial numbers, MAC addresses, whatever?
This is probably very much a - just because you might be able to - doesnât mean itâs a very good idea scenarios. In the states, we also start one of these with, âHere, Hold my beer!â
IF you can overcome the technical hurdles (and Iâm sketchy on if its possible, the clone should get it all in theory but I havenât finished peeling back the layers down to the radios) If theyâre ever ON at the same time - Stuff just gets very sideways. I also donât know if it gets sideways in a way that will require you to rejoin devices, itâll probably depend on your devices - but Iâd put money on yes. Especially if any of those devices are labeled Tuya or Xiaomi. You probably already know which trouble devices you may need to deal with - Thier stability will most likely get orders of magnitude worse.
After failover happens. lets say everything is âgoodâ it will look like a temp usb disconnect. In theory it should recover but - ok - maybe? Youâre way off in the weeds.
This is one of those where you look at the reason you might do a failover in the first place - uninterrupted operations for critical failover (failover because youâre not going to run flopping this back and forth across nodes all day based on capacity, this is an emergency keep my stuff running while we figure out what happened ability) - great! but youâd be sacrificing production stability for the possibility of retaining function in an emergency state to do it⌠Not so great.
Go with the network coordinator and get that working before you even try this.
There is some extra data on the stick other than euid. It can be backed up and restored, so is at least potentially scriptable.
Two powered sticks with the same euid will conflict, even if the backup stick is not actively being used. Youâd have to have the usb power leads on some kind of relay so two sticks would not be powered at the same time. You couldnât rely HA being down being the same as the stick being down.
On top of that you might also need to keep the backup stick up-to-date with the changes that might occur on the active stick.
It will be a management hell.
If you want that scenario that much, then start to look into Matter and follow that development.
Matter comes standard with the ability to do this.