There’s all kinds of fun stuff. Lightbulbs, smart sockets, temperature/humidity sensors, radiatorvalves, waterleak and fire alarms…
Start here: https://sonoff.tech/
See you in 3 days 
2 Likes
Thats the brand I am getting. That is what seeed had
Is that antenna supported by zigbee2mqtt ?
Huh
Antenna? Did I mention antenna somewhere my mistake?
The zigbee on board the yellow doesn’t have an external antenna if thats what your asking. So if the onboard antenna is not good enough then I have to make sure I have enough zigbee devices close enough to relay
I mean the internal antenna.
Because this is why I said see you in 3 days: Unless you want to be locked in by some vendor, you will have to setup your zigbee bridge: zigbee2mqtt
And not all chipsets are equally well supported
I also got a couple of zigbee microcontroller boards.
I thought that the home assistant s/w drove the radio controller on the board using different standards. Also others in the forum talking about it never mentioned it locked you into a vendor. But rather sometimes you just need a zigbee device thats powered close by. zigbee protocol I always understood worked by relaying messages through the zigbee net till the final device got the data.
Siliconlabs though, from my understanding is well supported with their range of chips
There is a native homeassistant plugin but it’s rather limited.
Zigbee2mqtt is the gold standard.
Any suggestions of where to get one
Anyhow I am off to see if they have made the sharks with lasers on their heads i suggested or if they get just a laser engraver
1 Like
Is it hassio you are using ?
I’m a fan of there Dutch mad man frenck it was where I fist came across influx and Grafana
Is there anything as energy efficient as Apple’s silicon? My focus will be on nodes per watt. Unfortunately I assume that will mean a Mac.
maybe thin clients ?
1 Like
It’s possible to build a Linux distro to replace PI OS on the PI 5 and mod the scheduler to have ACPI tasks be more aggressive in periods of minimal activity to reduce power consumption to 15-18W or less running a top hat with 4X NVME with decent speed over PCIe gen2 lanes.
This video is well worth the watch as its pretty current 4 April 2024 1.6M views
If you really want to keep away from power draw from a Utility company then maybe a solar setup using a nominal 12V system with a charger for backup.
A 100 to 200 Watt solar panel into a charger for the battery supplying a RPi or other SBC that draws like 5-20 watts will mean it doesn’t require power from a Utility’s grid. This should also keep the SBC running for a day or two if there is no sun at all.
Then of course a backup charger from the household power in case of bad sun days.
Battery size would need to be on the order of 20W * 2 days = 960 Watt Hours to supply a SBC running at max. Most good ones run at 7-10 watts with a SSD which means this would run 2 such units.
1 Like
I’ve been building small wind turbines for years, if you have wind then best to double up with two trickle chargers, one wind with own controller and one solar with own controller, save the trickle charge generated to two separate 12V battery banks,
then combine draw from both banks in parallel into a pony panel prior to the converter into the house., running it as prime and using the grid as backup if you can get away with such a setup to serve your daily demand, providing its not to large vs what you can generate and trickle charge. The converter should do the work switching from off-grid to the public grid when the batteries are low, and only switch back to batteries when they are 80% fully charged by wind and solar or double solar setups.
otherwise if you have no or crap wind at a reasonable height with 360 degree access and no canopy (trees or building rooves) obstructions within 100 yards being 10 yards below the wind turbine mount height,
then in the ‘no/crap wind’ case it’s best run two separate solar banks each with own controller to two separate battery banks wired as above into the house.
Redflow in Oz
https://redflow.com/
has some really interesting hot weather utility industrial strength battery storage which is Zinc Bromine plating tech, which works well in super hot weather…
We have one ZincAir player in the GWN (Canaduh) Zinc8 Energy now called Abound Energy doing similar starting to break through after three editions of crappy C-suite management… over 16 years , they got lured to a crap factory site in upper state New York…not, sure how that is going, given the R&D guys are in Vancouver, BC.
the actual original Zinc8 tech leaders/inventors doing the work are now running the Abound Energy company from what I can tell after this latest re-org , the patent is really solid Abound Energy | LinkedIn
Imo Zinc based batteries will dominate home off grid in the future at much lower cost per Kilowatt hour (even though Lithium has a 50% energy density advantage) as Zinc is everywhere in most all countries and is really cheap to extract, process, and the recycling industry is really good at reclaiming it…
2 Likes
If some one has the motivation then definitely wind is great.
I live in QLD and sometimes forget how bad sun is in some places. Here a nominal 50W panel will give you 50 watts (or more) for 8 or more hours a day, winter or summer.
Guess I was trying to be real simple with the setup and by going to 100-200W tried to suit most places where here I would choose the 100W and UK would choose at least the 150W or more.
Simple as for someone wanting to try it. Buy a panel package for a camper van (or similar) that has all the electronics and just put a battery onto it. Shove the panel in a sunny spot and tie it down with the attachments/pegs/whatever they give you and run the wire to the battery and SBC running off nominal 12 V
Wind requires more work to get going, so starting off with a simple “campervan” solar product gets them started with minimal effort. And then they can progress to wind if they want to.
I bought a few years ago a neat fold out solar panel to charge USB devices a 10 and 20W and thinking how easy that was and so portable. Not big at 10 or 20 watts, can even hang off a backpack (from the top to lay on the back of the bag)
2 Likes
Hey guys!
As promised I’m looking into contributing in a more decentralized way. Moving away from Hetzner and setting up a bunch of home nodes. I’m looking for some hardware advice as I’m sure there a lot more technical guys here than I will ever be.
I currently have 2GB up and down internet connection which can be used to some extend, but should not interfere too much with any home use like streaming as I’d like to prevent a divorce 
Depending on the profitability of providing resources I could have a dedicated 5GB up and down line for just running nodes. I’d like to keep that in mind when deciding on hardware choice.
I’d prefer hardware like mini pc’s over dedicated workstations, especially now during the winter months they can provide some heat throughout the house, making the heat feel less like a total waste. Personally I was exploring the route of multiple mini PC’s that have 2 TB NVMe and can run 50 nodes each but it feels quite inefficient / less scalable. Now I’ve also been told that it’s possible allocating NAS space to my local device, making it possible to run nodes directly on NAS. I wonder how a NAS using HDDS would deal with so many nodes accessing data. Does anyone have experience with this? Especially when the network is live, nodes fill & activity increases I’m not sure if HDDs can deal with the i/o required.
As mentioned, I’m not at all that technical, learning every day though. Perhaps I’m worried about non-existing issues, but figured I’d reach out to the community to see if anyone has any experience/knowledge they can share.
3 Likes
NAS will have too weak CPU to run more than 1 or 2 nodes on it, but you can have PC’s running nodes that have small HDD or no HDD and use NAS for storage.
There are many ways to do this and performance will be highly dependent on protocol you would use for accessing drives over network (SMB, iSCSI, SSHFS,…).
I was running such setup for a while, but went back to local drives and computers running independent. Performance was good enough, but it was more work to manage configurations and it was more prone to problems. You cannot reboot NAS or do almost anything with the network also switching of all client PC’s and so on.
3 Likes
Yeah, totally agree. I may have been a bit unclear in what I was trying to do. So I definitely do not want the NAS to run nodes directly. I’m looking for dedicated devices with a decent CPU. The issue I’m running into is that the mini PCs usually have the CPU to run 100-200 nodes, and RAM ranging between 16 and 64 GB. The NVMe storage is usually the limiting factor (either not enough storage or simply way to expensive to be worth it). So in my ideal world, I’d have a NAS with 4, maybe 6 HDDs. And several machine running 50 to 100 nodes each, using the NAS for their storage.
The question I’m having is, would a NAS with several HDDs get bottlenecked by the input/ouput capacity once the network starts filling / get used more often?
I’ve also read about NAS with HDDs and use SSD as cache for chunks that get used more often. Perhaps that is a valid option?
1 Like
@Mightyfool Also head over to the advanced router section. You might find your ISP supplied router will be the limiting force to having good internet with a couple of hundred nodes or more.
Also consider multiple IP addresses for you setup, and that way you can have multiple cheaper routers to give cleaner internet. Like a router for all your home stuff and small number (maybe 200) of nodes. The mikrotik 5009 is great for that.
Then another IP address (or 2) for the main body of your nodes. If you allocate 1Gbps to 1.5 Gbps to this second router (plan it out and set limit in router)
If you want to run 200 nodes on the one machine then realise it needs the memory, prob at least 16GB for when nodes grow in memory usage (at various times) and the CPU probably wants 4/8 cores/threads as a minimum. 4 cores 4 threads seem to limit to about 30-60 nodes. Maybe less
NAS as the backend storage prob would be OK since they should be able to do 100MB/s across your network and hopefully you have 2.5GB minimum for your local network.
1 Like
Current latest setup for safe node hardware at Location A that I have setup is as follows:
13 super old hosts in a cluster composing off:
148 cores
3.3TB RAM
208TB Raw Storage
13 hosts running Ceph Monitors daemon
13 hosts running Ceph Managers daemon
2 hosts running Ceph Metadata daemon
13 x 16TB (refurbished) disks split across 8 hosts (plenty of empty bays for future capacity increase if needed). Eventually each physical host will contain 1 or more 16TB disks.
Each physical host is linked with 4x1Gbps LACP aggregation (currently working on upgrading the interconnects to be 2x10Gbps fiber between all hosts).
All safenodes are run in an Alpine LXC (provisioned via automation).
They all have access to a dedicated ceph file system mount point. Folders are carved out at that mountpoint say for example as /mnt/safenode_vaults/<hostname>/<safenodeServiceX>/.
Logs are setup to a temporary RAM file system (reduces disk i/o & cpu cycles), and auto cleaned up when rotated.
At no point is data stored on a single computer’s local hard drive. Any additional shell scripts for further maintenance are on the Ceph FS system so easily accessible via every LXC if required.
Hoping for a high uptime, healthy nodes, and allowing rolling reboots for patching, if necessary with minimal downtime.
Planning on cloning this setup with Location B, C, … and so on. Definitely not sold on the non rotational disks such as SSD and beyond due to wear out even though they have started coming out with very high density capacity but the cost is still pretty high. It would require a ton of pci-e lanes to service all these newer interfaces. For now, I rather stick to used old and refurbished parts and spend money sparingly as and where absolutely required.
–
For monitoring, will be making use of influx db, telegraf, and grafana to monitor multiple locations all via single dashboard.
The influx db setup is HA as well with influx db proxy relays writing to 2 or more influx db engines (time series data), all backed by a different ceph fs filesystem as well (HA), which is also wrapped in front with a nginx load balancer.
Grafana at any point can query 1 or more off the nginx load balancers to retrieve the data as well. Allows for rolling patches as well.
11 Likes