Glad it has worked out for you! I still run my home on Hubitat, but I do have HA running on a RPi 4/4GB booting off of a 128GB SATA SSD. I am pretty impressed at how well it does discovering things on my home network, like 4 AppleTVs, 3 Logitech Harmony Hubs, Philips Hue bridge, Lutron Caseta SmartBridge Pro, printers, etc… I do feel it is weaker than other hubs in terms of Zigbee/Z-Wave support and ease of building Automations. It’s really the automations where it seems one has to geek out and pick a strategy… at least there are multiple options to choose from!
If any of you are Arduino maker-types, check out ESPHome. It is really pretty amazing, especially at how easily it integrates with HA. No C/C++ programming required!
I can only imagine the pain because you have WAY more devices then I do, and even for me just the process of re-paring all the devices and digging out the old manuals on how to do that was the most annoying part of moving. In the end, I’m sure it will be worth it though.
Totally agree, its gotten “easier”, but definitely not easy. They have come a long way though.
Home Assistant really seems customized to run on the PI, but I kind of like the idea of using a desktop or NUC. I think you get better performance and power. Definitely need the SSD with the amount of devices you have, as the SD card would crash quickly.
For additional performance, I’d also take a look at using the MariaDB addon. Its a quicker database versus the SQLlite that comes with home assistant. With a lot of devices, people have said it makes a big difference in performance - https://www.youtube.com/watch?v=FbFyqQ3He7M
That was a little tougher for me with the docker install - I had to do a cron job in linux, but this is where the fully supported hassos with add-ons works very well.
I know this threw me off a little too, and just kind of how people are like “why would you want that?”. I guess its more of a Smarrthigns thing. I created the “input select” helper and then created Home, Night, and Away. A lot of my Node Red Automations run conditions based on that “input select” mode. Its also a badge at the top of my lovelace page.
I’m definitely on team NodeRed - I use it for almost all my automations. The Home Assistant automations do run some things I have but anything more complex is definitely on Node Red - like leaving, arriving home, morning, and night changes.
I’ve already been peripherally looking at that Ogie, I’ve got my eyes on a modification to an Ikea standing desk to ‘smarten it up’ (there’s some light docs out in the ether from someone who did it.) and then I’m going to take that and apply the learning to my new couch and loveseat. (yeah I splurged and got those media couches… all the buttons!) I’m planning on setting a scene for the couch to match ‘watch Star Wars’ Lights go down, D+ starts up, deep links down to Ep.4 and the couch reclines just as the opening scroller… well, you get the point.
I have moved all but Zigbee devices and 1 scene off SmartThings now over the last 2 months. My setup:
10 year old Gateway PC with an ancient Intel i3, 6GB RAM, and 1 TB HD. Running Manjaro Linux with HA, Zwavejs2mqtt, and Kerberos running in Docker. The cam software (Kerberos) isn’t setup yet, but will be soon.
Went with Manjaro over Ubuntu because every update of Ubuntu gets slower and slower on an old laptop I run it on…or maybe it’s because the laptop is 13 years old!
Z-stick is a new 700 Series Zooz
Zigbee coordinator is TBD. Gotta find the perfect match for my 30 or so Eaton Halo recessed LEDs and also must play well with Sengled.
Everything is stupid fast in HA. Automations are a learning curve, but nothing I haven’t gone through before coming from Vera originally. Scenes, Scripts, just understanding what they all do is a task.
Looking forward to finishing the move.
Also, going to install Multi System Reactor in a Docker container. Won’t find it in the forum yet, but version 1.0 was recently released. It will take the place of the logic engine. It’s an alternative to Node-Red with it’s popularity coming from Vera.
Well - we tested BOTH methods of moving to SSD tonight. I originally tried to clone the SD - it got hosed somewhere in the reboot and - we lost the patient.
Only to ressurrect her by using the snapshot I took immediately beforehand.
Preferred recovery method is definitely:
Dump new OS image on the M.2 (the case let me use it as an external HDD by plugging in a USB3 A<-> cable!) (don’t forget to ssh)
Use SSH get in - set options including pass and make SSH permanent (reboot)
Run official HA supervised install script (update > Docker > Supervisor)
Login (create dummy account - we’re blowing it away anyhow)
Install Google Drive Backup Add-in to supervisor,
download snapshot
restore (FULL) from snapshot.
Glad you were able to recover, a crashed system off the bat would not have been a good first impression. At least you had the backups which I think a lot of people often neglect.
So, the main issue here is that Home Assistant has a “recorder” integration, that is responsible for keeping all your device’s event history. Its defaults are to commit to disk all the states of your devices every second, and keep 10 days worth of history data. There is absolutely no way an SD card can handle that kind of read/write rate for even a modest amount of devices, yet many people use a PI with an SD card as their initial hardware for setup. Then, the result is as expected, and many people will experience a crash.
The documentation in the recorder integration specifically warns against this - Recorder - Home Assistant and advises to change the default to 30 second commit times if using an SD card. However, that requires changes to the config.yaml file, and most people learning this new system will not figure this out or read this documentation until it’s too late.
It almost seems figuring the recorder out is a “right of passage” to having a successful setup for new users. I’ve had an SSD the whole time so was spared this issue fortunately. I don’t know if or how it could be handled better, and I feel best to stay out of that debate. Being an open source project, a lot of times criticism is taken as “we worked hard on this, you didn’t pay for it, didn’t you read the warning in the documentation? If you don’t like it, submit a PR and fix it.”
There’s a post I read early on that I feel sums up the situation of the whole Home Assistant project as a comparison to a “community park”. Its a good read and I feels gives some perspective on the project. Leaving Home Assistant, not worth the headaches! - #22 by 123 I am not a Python coder and my limited Java knowledge doesn’t help much here in terms of contributing to the project, but I try and help where I can by responding in the community and attempting to stay positive, or commenting on a proposed Github change that could impact my setup. Wink and Smarrthings (especially as of late) have set the bar very low for me, and I’m a huge fan of Home Assistant and rarely can find anything to get too worked up about. My boss always jokes “everyone should have a bad first wife”, and now I’ve had two! Still, just like everything, there’s always room for improvement.
There are High Endurance microSD cards that will handle a much higher volume of writes versus a typical microSD card. These cards are often used in always-on video cameras, like dashboard cameras. I use Samsung Pro Endurance SD cards in all of my Wyze cameras around the outside of my house, which are all configured to record continuously. I have not had one of these cards fail yet, whereas other ‘traditional’ SD cards did all fail eventually. (Yes, I know that they will eventually die, as will all flash memory.)
Thus, if one is going to use a microSD card to run a RPi, I believe they should choose a high endurance model, with a very high capacity (to take advantage of wear-leveling over all of the flash memory cells.) Otherwise, it is best to boot and run the RPi from a SSD.
If I were ever to migrate my home over to Home Assistant, I think I would simply run it on a NUC, or similar hardware, just for the added flexibility of having extra horsepower and storage available for running other apps, like InfluxDB, Grafana, etc…
The issue is that if you use just a class 2 SD card, and leave the recorder at its default 1 second read/write and don’t change it to 30 seconds, it’s not a matter of if your SD card will fail, but when. It kind of sets you up for failure and just throwing it out as a warning for a new user. I think the latest users jumping ship from Smartthings are people with a lot of devices due to the newly imposed 200 device limit, so this could definitely impact them if they are reading this thread.
Totally agree, especially with 200+ devices. Here’s another good video on hardware recommendations and comparisons. https://www.youtube.com/watch?v=rnh7vzOWABU&t=233s I like his idea of getting a surplus used office desktop. You can pick those up for a pretty good price and get some impressive hardware specs.
fwiw: I use the alarm integration for my nodes.
Armed_home and armed_away is enough info for me.
Ive connected presence detection to it and works pretty well.
Nodered is very very powerful, biggest turnoff is the fact you can’t do it on the phone. To me there is no shame in mixing the automations
The end goal should be to eventually remove your SmartThings hub completely. I have tons of beginner guides on my site, but specifically in this post I have guides for moving a bunch of SmartThings-branded devices (motion sensors, buttons, door sensors, etc) here: https://smarthomepursuits.com/how-to-migrate-from-smartthings-to-home-assistant/
I started with an Rpi 3B+, and that has turned out to be far slower than I expected, especially when I was restarting HA. So I’ve made a backup, and made a Supervised installation with Debian underneath on a HP T630 Thin Client. It is far more faster and has 16GB of memory, plus an M2 SSD slot. Plus it is not actively cooled, so I don’t have to worry about noise neither.
For Zigbee, I started with a CC2531 dongle, but that has turned out to be too weak. I haven’t really had any problems, but once I started to move more devices I experienced some slow down on response times. So I upgraded to an Egony CC2652P dongle which does the job. I had some issues with that for a few days, due to some remnants in the memory of the device, but I reflashed and erased that part and now it works as expected.
@Nameless, look at the list on Z2M for the dongles, there are plenty good dongles and “gateways”, the market lately’s flooded with CC2652P solutions.
@ogiewon, I would disagree with you regarding the endurance cards. I do use them, but their wear capability was really designed for continuous recording of bigger chunks of data, not for small file access. I believe that makes a big difference. But I do use them in cameras and in one of my RPis as well. Failure of an SD cards is always imminent. It doesn’t mattet what kind of it, it is just the question when.
@NathanCu, talking about the 20’ ladder. I started to move my Zwave inwall relays to HA the other day. I removed the switch cover and switches from their place, and had to realise that 3 relays are not there where I expected (Roller shutter controllers), so I still have to pull out the ladder to climb up to the attic where they must have been installed (by me).
Yes I had the same thing happen and had to upgrade it, and I don’t even have that many zigbee devices. The supported adapter list you mentioned - | Zigbee2MQTT has grown with new ones even from when I bought my new adapter a few months ago. I went with the first one listed that’s the Electrolama zig-a-zig-ah! (zzh!) and bought it from Tindie. There was a wait list, but it was shipped relatively quickly to me in the US from the UK. It appears in stock now, and Its made a huge difference and I would recommend it. It’s one of the CC2652P chips you mentioned and probably any of the devices that use that will be a huge improvement over the CC2531. The CC2531 specifically says “not recommended” now on the zigbee2mqtt website. Here’s the link to order one or check it out
Honestly, No - I had managed to get most of my stuff to run locally before I moved to HA. I will say the stuff i do have running so far is much faster. (Except the blinds in the front of the house, which I suspect is a weak spot in my ZWave mesh…)
Right now I’m honestly still just dealing with the pain of moving at the moment. It takes time to move that many ‘things.’ So far, I’ve got about 95% of the hardware moved and I’ve started rebuilding the automations. I’ve still got at least a week of work before everything is back to ‘parity.’
Edit: Oh - and don’t need Alarmo - the DSC panel integration gave me everything I needed by default. I don’t need to overlay anything on top of it.
Someone recently pointed out some issues in the way zwavejs heals the zwave network and updates neighbors.
The full network 'heal" can actually temporarily make things worse. The network usually figures itself out in a couple days though. I’ve found healing individual nodes that are slow works better then healing the whole network
A bit late in saying hi on this thread given I moved over to HA in August, but I made it… HI Y’ALL (not Texan but live in TX so I adopted their lingo) EX-ST USERS!! Good to see you here!
One issue I had with ST, and/or ST forums, is that it was not geared to be too tech detailed. I struggled for years with excessive zwave traffic and related issues. After just a few months on HA I am learning WHY I was having those issues and WHAT I can do to mitigate them. I found some posts on here that were way more helpful than “you have too many devices” which pointed me in the right direction.
A couple critical things I learned:
Better to go insecure than use S0 as the latter increases 2 to 3x the traffic.
Get rid of all old zwave (non plus) devices as they will bring down the mesh network speed. If this is true, then I have one big issue… I have 3 non plus zwave locks that would require about $600 to replace. I am currently unemployed so not a good time to make the upgrade, and anyhow, I would either want a 700 series lock or a Matter lock… given how rarely those would get replaced. Neither exist as far as I know so… what do I do? Any suggestions?
The above is basic stuff but even ST themselves never suggested it as a mitigation strategy. All they told me was to tone down Zooz motion sensors, which I did, but my issues remained. Also, the funny thing is that I always started having issues with my mesh once I got to including my locks. Since I work my way from hub out, that usually happened at 40 or so devices. Perfect storm… # of devices and 3 ancient slow locks!