1. The simplest would probably be having the Tellstick only as a backup solution, to turn off power to the USB over TCPIP box and turn it on again and then have EG reboot the VM, which always fixes the connection. The RFX would turn off the USB over TCPIP box for the Tellstick then EG would reboot the VM when that is lost, so they sort of take care of each other.
2. A bit more complicated version would be to have a fully redundant system on one VM with the above setup in addition to a separate heater script where the Tellstick also receives the temperatures from the sensors and also turns off and on heaters. The script would be almost identical, only with a two second wait before sending to the heater. That way I avoid the signals from the two units interfering with each other.
3. The most complicated but probably most secure version would be to have two separate VMs, running like the system above. One with the RFXtrx and the USB-UIRT running as the main system and the redundant one with the Tellstick both doing exactly the same job (with two seconds delay). No problem to set it up, and I would assume that it would be more secure since a crash in EG on the main system could be caught (for instance with a function that listens for commands from the RFXtrx and does the off/on and reboot dance if there are nothing sendt for so and so long).
What would you guys choose from this? Stability is paramount, I'm just not sure if there will be much stability to gain from complicating things too much (especially since I haven't had crashes in EG or Girder or ONH for a long as I can remember except for when I'm changing stuff in the system).
What I won't do is to add more systems (hardware or software) to the settings, so using a Raspberry Pi or something isn't really an option in this.
