I ended up splitting the load between my new C8 and old C7, as I was having memory issues after a couple weeks on the C8.
Moved my LAN Integrations (Kasa, SwitchBot, ESPhome) to the C7 and share them back to the C8 via Hub Mesh and it's been working fine.
Current c8. This usually holds for a month or so but then I get a platform update and have to reboot. Never had the patience to just wait.....
With just one addition...
WHATEVER level of component quality, reliability, and extremes tolerance that went into the C5 ...keep striving for that.
We're into the third winter/summer (freeze/simmer) seasonal cycle of unit survival in dusty unconditioned outbuildings. Not everybody can put these in nice conditioned home/office spaces. With the exception of water I dare say they've performed to MILSPEC even with the dust.
This is exactly what I'm doing. I've moved Kasa, Ecobee, LifX and the "Precipitation and Weather Monitor for NWS Data" device over to my old C-7. Hoping this is enough to buy me more time between boots. If necessary I'll move Echo Speaks over too. It's just such a touchy beast, I'm not in a hurry to tempt fate.
You've been managing multiple hubs so long you forgot how much easier one is. 
I get what people say about CPU use. That was never a problem with my C-7 running everything. It ran out of memory and needed to reboot. Failing better tools to figure out which app was responsible, I've split the load. It is what it is. Just saying if I could spend $350 on a hub that has so much headroom so as not worry about resources, I would do it. I like the idea of a built in battery back up. But I have my network gear and (NOW TWO) HE's on UPSs so that's fine. We recently were without power for about an hour and a half and I never lost network access including HE. Of course she didn't have anything to control. Except my power failure notification solution which worked just fine.
I have found that typically Zwave Zigbee stuff is very resource efficient. It all generally goes down hill once LAN stuff starts to get involved. For some stuff that is inevitable, but It goes back to what i have heard over and over. Stick with Zwave/Zigbee devices when possible.
If I was going to break my setup out the very first thing I would do is break out the Lan/Wifi integration devices to their own hub.
As far as the conversation about more resources, I see both sides to it. Simply put when things are kept simple yea we don't need much, but once things start to add up and get complicated it doesn't take much to overwhelm that small system.
Having the ability to scale it with your personal environment isn't a bad idea. I would love to load Hubitat as a docker and just allocate resources to it as needed, but i understand why they don't allow it as well It is just another variable. Also being in IT i have seen how the saying "if you build it they will come" applies here. Making the machine bigger could likely not do anything for peoples usage issues. Maybe postpone it a little as some have said, but it could also promote bad habits as it just obfuscates them.
That thread is also about the subject of capturing extended logs with advanced filtering and such. It works great and the INfluxDB Syslog Device driver to load INfluxDB seems to work well with low system impact.
See the things is, I guarantee that whatever is causing the memory loss would keep chewing up memory even if you had 2 tb of memory in the thing
To some extent I agree, but there are legit reasons a app may take up more memory space. The Govee Integration is a good example. To enable scenes I am having to store the values for each scene command in the app to be used by the devices. This has added up considerably when you consider the quantity of devices. I have taken steps to reduce it as much as possible and am thinking about more ideas to further trim it down, but no mater how you look at it advanced features like that are more then a small on/off command attribute or something like that could benefit.
I also think the internal processes are getting to the point they may benefit from it as well. Most of the time I have noticed a hit in responsiveness has been accompanied by a sudden drop in memory related to DB access. Yes sometimes it has been something else behind the scenes, but it has also been related at times to the DB backup. I have seen my hub take a 80MB hit just by doing a backup.
I think something else to consider in this discussion is simply what are we talking about in actually price impact. I don't know what memory is actually in the unit, but when I look at Micron FBGA on Mouser the difference from 1GB to 2GB chips isn't that much money. They could probably quadruple the memory and only charge another $25 bucks. That doesn't account though for if the Processor in the unit can take that much more. If a upgrade is needed on it, that may make the difference bigger.
Ofcourse I say all of that but I am seeing mixed reports of weather the S905x in the more current HE's can support more then 2GB of ram. If they had to go to a new CPU, pricing could get more complicated. We also don't know how much ram is in the box and used for OS vs JVM functions. We know we see the JVM free memory is around 600MB after a restart but other than that we don't know.
I don't suffer hugely from free memory problems but I've never seen it that high after reboot. This is mine after last reboot on July 8th:
It's now down to 208 MB after 40 days
That is not very surprising. It depends on your environment. I see 600MB + on my dev hub that has limited stuff. My Prod hub is similar to what you are showing there. you can kind of see that in the below image.
In regards to this suggestion (of a "SUPER HUB"), I must admit that I'm of two minds.
First of all, from my long history in IT, we all have used to the approach of "throwing improved/upgraded hardware at a problem". This was a very common technique, and it worked very well in many circumstances. The HW got much less expensive and much more powerful, and the cost of SW adjustments (programming/debugging) got much more expensive. However, some times this approach didn't work. In some cases, due to user error/poor programming/etc. no matter how much HW we would throw at a problem, it wasn't enough.
To be frank, this was a lazy man's approach.
The other approach was to spend the time necessary to uncover the root of the problem, and properly debug the issue. Sure, there was the very infrequent case of Operating System bug, or unusual confluence of circumstances. However, in general, many of the cases that we used to solve by throwing HW at the problem could have been solved by smarter/better programming.
This approach took MUCH greater time and effort, and may have cost more in various circumstances.
Theoretics aside, in this particular case, we all know that the Hubitat box is not an all powerful box able to handle all workloads thrown at it. However, we all know, that the box can (and does) handle 95%(? guess - maybe more) of all general Home Automation environments and situations. Furthermore, they made the box even extendable with the use of Hub Mesh so that you could easily leverage more than one box.
It's a compromise situation, and in my humble opinion, Hubitat has chosen the right balance.
One of my all-time favorites!
Right up until I realized he was a racist piece of garbage..
I think he's referring to the cartoon, not the cartoonist.
At that point same thing... Though i don't wanna devolve the thread so I'll step off my soapbox...
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.
Download the Hubitat app
