Home Assistant and NAS

Hi,

I’m currently running a little Ubuntu server with HAOS in a VM, and Plex server as a snap. Additionally, the server provides a RAID-1 set of 2 HDDs via Samba. However, I’m not happy with this setting, mainly because I want to reduce energy consumption (which is currently around 35W when idle).

One thing I’m thinking about is using a HA Yellow for this purpose: I could run Plex server as an add-on, use an external RAID-1 case for the HDDs, and attach that case via USB3 to the Yellow. The remaining problem appears to be how to share my HDDs via Samba. Is this possible? Does anybody have any experience with such a setting?

The second option would be to build a Pi based system from scratch, and run HA as a Docker container on that system. This would most likely allow for all I’d like to do and give me more flexibility, but the configuration effort would be considerably higher, and I’d like to keep things simple… Are there any tutorials for setups such as this one?

Finally, I could go for a commercial NAS which allows to run HA on it (probably as a Docker container) This is probably the most simple, but also by far most expensive solution, I would guess. Do you have any hardware recommendations? I have 3.5’ HDDs, I’d prefer to use 4 HDDs with RAID-5 in the long run, but 2 HDDs would probably also be fine since these systems appear to be much cheaper…

Looking forward to reading your recommendations! After all, I think that this should be a rather common use case…

Best
Christian

I run Home Assistant OS virtualized on my Synology DS220+. Synology offers Docker and Virtual Machines and a lot of other nice software. I run on my machine:

  • HA OS on virtual machine
  • Various Docker containers
  • Back up solution for all household PCs and phones
  • Video streaming
  • Photo cloud
  • NVR

If you want 2-bay, my model is the way to go. There are 4-bay options, but as you already mentioned with a higher price tag. The only downside was the low RAM specs, but there is the possibility to add more very cheap.

After some more research, this indeed seems to be one of the most promising approaches… Can you elaborate a bit on the memory issue? I think that 4GB are recommended for running HA, and the DS220+ apparently supports a maximum of 6GB RAM - How do you run HA and then still some Docker containers? And did you do the RAM upgrade yourself using standard RAM? Thanks…

I run with 2GB of 8GB allocated to the HA VM on a D918+ and it runs fine, you can always adjust this as needed.

I used this one:

Together with the built in RAM I have 10 GB, which is more than enough. You could even build in more, although it’s not “officially” supported.

I allocated 2 GB to HA OS.

A NAS will not be any lower than this in power consumption. I have a Synology DS1515+ running HA, Plex, and a number of other docker container programs (see my signature for a link to my current setup). This is a 5-bay with disks in every slot running an SHR disk pool (synology’s version of RAID-5). I have the NAS and 1 external hard drive on an APC BR1300 UPS and get a runtime of ~45 minutes during an outage. That translates to a roughly 7% load or about 130W. Anyway, my point is, there are way more economical devices in terms of power consumption vs price that can run everything you want, but at the cost of performance.

It’s great to be energy conscious. But there comes a point where the performance or demand of a system requires you to look beyond that and get the hardware necessary to do the job.

Those are the energy comsumption specs of the 220+ and they are realistic:


Add a few watts for VM and containers. You can operate it fully loaded at 20W.

My whole setup containing

  • NAS (with all the services mentioned above)
  • 650W UPS
  • Router
  • 2 Switches
  • 3 Poe UniFi APs
  • 1 PoE Camera

draws 45W

That’s actually rather close to my setup (despite the UPS and the camera), and that’s the numbers I’m looking for :slight_smile: Have ordered a DS220+ (and 16GB of additional RAM - I figured that if I increase RAM above the supported amount, it doesn’t really matter whether it’s 2GB or 10GB too much - we’ll see :slight_smile: ).

One question, though: HA logs a lot of data to its internal database (and I’m writing to InfluxDB in parallel), so I guess that if I run HA on the NAS, it will barely be able to hibernate the disks, if at all. Is there a solution to this? I thought about attaching a (permanent) external SSD via USB, and to configure HA such that it’s writing to that drive (and e.g. backup it to the mirrored NAS drives nightly)…

Hibernation is very unlikely. Since my setup includes Surveillance Station which writes 24/7 anyway I didn’t have to think in that direction.
But is it really worth it? You might also have other services (HA add ons/docker containers/Synology packages) that require to read/write. Outsourcing all DBs and configurations might be quite some work with few benefit. External DBs also won’t be included in VM snapshots.

That’s true - here’s the power consumption of my NAS so far (with HA running):

image

Pretty nice! :slight_smile:

I’m also running on a DS220+ with 2×4TB drives that I upgraded to 6GB RAM. HA and other stuff (Plex, PiHole, Mosquitto, AppDaemon, Grafana, Influx, MotionEye, ESPHome, memcached and leaf2mqtt) all running in Docker containers, currently showing 41% RAM usage. Performance is not an issue: all actions appear to happen instantly.

Drives never hibernate (wouldn’t due to MotionEye) and nor should they for best life and performance.

I went the weird way and equipped my DS220+ with a whopping 2+16GB of RAM - so far no problems whatsover. Performance indeed is great!

Concerning hibernating: Are you saing that NAS HDDs do not benefit from hibernating if it comes to length of life? If that’s true (I didn’t know that), then I’m not going to worry about that question any more (I guess the power savings from hibernating are neglible)…

There’s not much concrete data. My guess is that the extra wear and tear and thermal variation from the harder work spinning the drives up from rest is at least as significant as the extra steady-state wear from leaving them running — and you avoid the start-up delay if you keep them running.