Official stance on memory and hub slow down?

interesting on my c7 in nh these are all the cloud integrations

  1. honeywell tcc
  2. honeywell home
  3. kasa tplink
  4. tesla
  5. 3 different ecowitt weather stations in 3 differnt locations
    6 2 differnt web based weather ie darksky , and openweather
  6. color cast weather lamp which also communicates with openweather
  7. 3 different wireless tag integrations for tags in 3 locations
  8. amazon echo app integration
  9. konnected alarm panel integration
  10. 2 different harmony hub remotes through local net.

quite a lot so it looks like the c7 seems to be much more capable than the c4 or i am just lucky which cloud/net integraitons i am using?

1 Like

I was thinking this too. I have Honeywell, Kasa, Echo Speaks, and some others I think but can't recall

I think the key may be how often they poll the net/cloud vs how many.
You need to be carefull to not be checking over the net every 10 secs say. my therms for instance are only every 15 minutes. kasa only every 10 minutes or so. etc.
I think if you screw that up you can bring the hub down very quckly.

The konnected has no option for that as it is instanenous but at least it s totaly local and not over the cloud. Also, I avoided the kasa cloud version and kept everything local as well.

2 Likes

That's crazy and with no slowdowns. Maybe the C7 is better at LAN devices. I was too concerned to try when I moved from C4->C7 and left most LAN devices on C4

it could also be the latency and how your hub is connected. ie wifi vs hard wired..
mine is connected from the upstairs to downstairs via gocoax moca devices so basically hard wired in..

I kept it downstairs as that is the middle floor in middle of the house whereas my internet comes in on the upper floor.

I’d be willing to bet your initial windows machines have a bet well defined SOE that strips or all the crap MS includes with the OS?

I’m an ex server engineer (worked for IBM in the 90’s/2K’s and later HP ES up to 2013 as a TL of 13 Server engineers), so I know windows server is very stable, it’s the desktop versions I distrust,

The hilarious trend I’ve noticed since about 2012 is many many windows server engineers switching to MacBooks for their primary work machines. At the company I work for, most SRE engineers have MacBooks now despite 60 of our cloud services running on windows server with SQL databases.

I’ve found the same, I run a 2nd c7 for most of my LAN / Cloud integrations.

2 Likes

I have a couple integrations that the developers re worked to run on the hub instead of a bridge (I was using a Pi4). Funny thing is that a second hub is cheaper than a pi now! I WAS thinking about moving back to running a bridge but now I think I'll run a second hub

2 Likes

Well yes, of course these systems are all running well tuned Windows 10 image that I personally developed. Same image is deployed to all production systems, and these systems run 24 hours a day, 365 days a year. We do buy decent quality hardware for these, but not not more than about $600 per system.

End user computers are handled by a completely different team. However, they too are extremely reliable. All of our software engineers develop using Windows 10 laptops and desktop systems. BSOD issues are extremely rare, and are usually caused by failing hardware. If users (whether home or corporate) spent as much money on quality Windows PC hardware as they do on Macintosh systems, instead of the cheapest thing they can buy, their systems would be much more stable and reliable. I am not a Windows fanboy, nor an Apple fanboy. Both systems have their pros need cons. I personally use an iPhone, iPad, and Apple Watch along with a Windows 10 Ultrabook and a Windows 11 AMD Ryzen Desktop system. All very stable and reliable.

4 Likes

my mail server running here is a tank .. an older dell r900 server with dual redundant power supplies. and 128gig of memory.. 4 6 core xeon processors and sas raid with 10 sas 10000k 2.5" drives with hot spare.

i have upgraded the network to 10gig and it also has a drac remote access controlle rto reboot power off on etc via a 2nd ip..

it is a pretty old machine and dell said it couldn't be updated past server 2012 but once i put a newer raid controlle ri could update to either server 2019 or 2022.

it is built like a tank.. had like 6 or 8 fans.. i did put quiter fans in as it is in my office.. it does use a bit of power and generate heat.. lol

3 Likes

I have a pretty large HA system and broke up mine by protocol many years back with C4s - coordinator, Zwave, Zigbee, and LAN. I believe it was early 2021 when I was having to reboot these C4s daily or every other day so I migrated them all to C7s. I am now going 3-4 weeks between reboots (depends on firmware updates). Until today they were all up since 4/28. Huge difference with hardware. My LAN hub was the easiest one to migrate to C7 because I didn’t have to re-pair anything just simply restored a backup from C4. Of course I had to update hub mush but that was easy to update the links.

I had a similar experience with my C-4s except I migrated my Zigbee stuff to a C-5 and the Z-Wave to a C-7. Had a spare C-5 so that became my "cloud/network" app hub. Since I handle most of the complexity outside of HE (using NR) my hubs free memory has remained fairly steady. I do have some sequences that will reboot the hub at a specified time if it dips below a threshold. So far so good though.

Interesting that Maker could potentially impact things.. as thats pretty much my main HE app.

This... Even cheap hardware runs fine (even though it physically falls apart if you breathe on it wrong) Windows for a long time now is largely creep free. What I notice the most is on aging systems is the hard drive failures (or rather pre failures). I get calls on "My system is so slow" and within 2 minutes I can confidently say the motor is on the way out. (provided it's not an issue with outlook's ost file approaching the 50 gig mark but that's a whole nother thing to get pissed off about). Usually replacing these hard drives with SSD's not only solves the overall problems, but extends good hardware for another couple of years. We still have production work stations running i5 3ghz cpu's with SSD's and Windows 10 that are 5-6 years old and they are very stable and quite speedy. As server drives fail, we also replace with SSD's for the same reasons.

1 Like

2.3.1 imo has worse memory leaking than previous releases, I used to see it take around 5 days for available ram to fall below 300k, now it’s under 2 days.

I wasn’t referring to BSOD’s, I was referring to apps etc inexplicably slowing down and acting up.

Microsoft has done a good job protecting the kernel since w2k3, it’s just all the other components that aren’t as well behaved.

Possibly. I haven't owned a Windows computer in about 20 years, but at about the same time I purchased an iBook for ~$1000, I purchased an HP laptop for about the same price. I don't remember what version of Windows it ran (I think Windows XP?), but I can recall at least one BSOD a week, while the iBook just ran. The only thing I used the HP for was to run a statistics package - SPSS, so could have also been the software I guess.

3 Likes

We’re like twins, except I have a Ryzen powered laptop. I have built my own desktops since the 90s, so I’ve always used windows. My always on pc that runs Homebridge is on Windows 10 home. It only gets rebooted for updates and has been super reliable. Maybe because it’s a barebones system; integrated graphics (skylake), one SSD, 2 sticks of ram, that’s it.

2 Likes

I run an i5-based Win10 'Home Server', with 16GB of RAM, a 128GB SSD (OS), and 3 x 8TB drives for storage. I run Homebridge, Node-RED, InfluxDB, Grafana, Channels DVR, and Plex Media Server on it 24x7 with no issues whatsoever. I have disabled the automatic patching and rebooting to prevent any interruptions to OTA DVR recording and media content streaming throughout the house. This has been working great for years now. I only reboot it about once a year when I patch it manually.

3 Likes

There was a change of how memory was displayed in the last couple 2.3.1 releases. You can't compare older versions to the current version memory stats. The release notes have at least a couple references to things that affect memory and memory reporting. Release 2.3.1 available

The free memory metric in 2.3.1.142 better reflects the state of affairs, but it is different from pre-2.3.1 and even different from early 2.3.1 public releases.

This also means that the "reboot hub when free memory goes below 250" approach is no longer applicable. You can let it run all the way until the "low memory" alert pops up at 80m.

9 Likes

Download the Hubitat app