Proxomox - Install HA Add-on or Install Standalone Docker container instead (in Debian VM/LXC) ? Your opinion in term of Performance/Load?


I was using till now HA on a Synology but decided to move to a dedicated x86 Plateform with SSD n which I have installed Proxmox + HA on it

I would like to have your opinion on best practice to install Mosquito, Doods, MariaDB, etc… is this better to install them through the HA Add-on or install them in an independent way under Docker (in a Debian VM or LXC for example)


I personally like using Docker. Probaby more habit at this point, but I really like how everything is setup currently. I run a Celeron based NUC with Proxmox. One VM is just Debian with Docker installed. Then I have Home Assistant core, ZwaveJS2MQTT, Mosquitto, etc. on there. I also run a few other LXC containers out of Proxmox for Unifi, Wireguard, OpenVPN, and PiHole. It runs great overall, even being a slow dated Celeron.

At the same time, since you’ll be running Proxmox, you could always try both. I know its not just plug and play, but you have some flexibility to go whatever direction you want and see what works for you.

EDIT: Typos

I run everything detached from Home Assistant in its own LXC container.

Screenshot 2021-02-10 at 8.08.23 AM

Having add-on fetaures organized like this is good in term of structure I think, but in term of CPU/RAM Load or simply in term of performance, do you think it’s better rather than simple HA integrated Add-on which means all containers located in the HA VM ?
As Gregg098 said, I could try witrh proxomox different settings to see how it reacts, but you already have perhaps a feedback to share on this


CPU/RAM load for LXC containers is lower than for a VM, your are “closer” to the host and the overhead is smaller.

1 Like

I find that LXC resources are around half of that used by a VM.

Another reason I like to run detached from HA is, if/when Home Assistant is down all of my critical automation’s keep working! I utilize Node Red, MQTT and Z2M in separate LXC’s. This allows me to shutdown, update or even try new things with HA and not affect the family.

I also like having each so called add-on backed up separately.


I should also mention that I run Home Assistant Container for home automation only.
I don’t need any fancy graphs, logbooks ect… so, it requires no database (recorder not loaded) therefore no constant writes to hard drive, corrupt .db or wasted resources. I take this approach throughout Home Assistant. I’ve ended up with a very stable, fast and enjoyable platform.

I used few of both CT and add-ons and avoid level of complexity with portainer dockers in HA.

  • detached linux containers (CT) for services which are not related to automation: nginx reverse proxy for example.

  • included add-ons in Home Assistant for services by design already very adherent to home automation and Home Assistant such as Nodered.

Since there are proxmox home assistant users here…I could use guidance. I have 4 core J5005 NUC with 8gb memory using my TV as it’s single display. I want to run home assistant supervised, pihole and debian 10 desktop. My operational system is now debian 10 with pihole and all the home assistant containers under docker. All is good…I use portainer to manage docker.

I am experimenting with Proxmox which asks how much memory and how many cores to give to each container. Can I assign multiple containers 4 cores and they share? Can I assign two containers 8GB or memory and they share?

Ultimately, my goal is to have the NUC running home assistant, pihole and also be my media server replacing roku for streaming using my TV as the only display for the NUC. Proxmox is so that I can have experimental containers without impacting the operational ones.