Yesterday noticed some strange noice from my INTEL NUC (fan?).
Then I started thinking what I do if the device will fail certain day…
Of cours I have regular back-ups, etc. So generally no problem…
But then finding new machine, installing again HA from back-up could take few days…
Maybe few hours if you good prepared…
And then I came to question - is it possible to set-up HA on 2 devices?
Working like some kind of ‘RAID 1’ system…
The purpose is that both systems are working all the time together, but if any of them will fail (hardware) then we still have the second working…
You’ll find many threads on “high availability” that have covered this in depth.
At a very high level trying to run them as active/active will cause problems if you have automations active on both, since they don’t know about the other. Plus all the challenges if you use Zigbee or Z-Wave.
My plan is to run HA OS in Proxmox, then off-load anything that can be, to either bare metal or an other Proxmox VM.
Proxmox can be set up in high availability (which also happens to be HA ), and VMs in Proxmox can be set up with canaries/heart beats to restart on fail, restore from known good.
My understanding is that MQTT brokers can be set up in all kinds of ways, eliminating the Zigbee & Z-Wave concerns.
Making the hardware high availability is of course just standard dual power supply, backup power, RAID of your choice (I like 10),
It is a balance between over-engineering and creating a maintenance nightmare and not getting enough 9’s.
Proxmox is a great environment, but the Zigbee and Z-Wave issues are not so easily dismissed. In most cases you would need to physically move the Zigbee or Z-Wave transmitter from the USB port on the old system to the new. For some use cases (such as the OP’s) this isn’t so bad. But for folks trying to set up high availability at, for example, a remote cabin, it’s a difficult problem.
I also did this brain exercise so interested in answers/solutions/ideas.
For zigbee:would a network connected/ethernet coordinator not solve that problem?
At least that’s a way to exclude that issue and Proxmox HA can take care of other stuff, no?
What about having 2 physical machines with HA, running simultaneously with the same config (somehow mirrored): what happens with automations.
Would that work?
Anything that touches cloud services will double your use of those, potentially putting you over usage limits
Automations will run on both platforms, duplicating everything and potentially causing problems (eg some devices when told to turn on when they’re already on will turn off)
I’d go search for the existing threads and read up, all this has been covered many times already
In a High Availability Model… not really. As far as reinstalling HA from a backup, I did it on my NUC in all of 30 minutes, Flash HA to NVME and upload the backup when HA boots and asks you to setup the username with ‘Restore from backup’
if you are simply looking to reduce the load on your system and split it between two systems so that two systems are sharing the load and not causing the systems to be under high load. You can create a master HA instance with as many slaves as you want.
This forwards entities from your slaves to your master, then you can write your automations and such on the master, and link your android or ios app to the master.
On the Zigbee/Z-Wave point that several folks raise, I would strongly recommend that you run both “services” in docker containers on separate servers. This should allow you to “listen” to both using MQTT.
On the same note I would also suggest that you run HA in container tusind docker compose. That greatly speed up recovery.
), and VMs in Proxmox can be set up with canaries/heart beats to restart on fail, restore from known good.