I've read through the two available threads without a real answer. I'm a new HE user. I bought a C7 to migrate to from SmartThings. I have about 80 physical devices scattered throughout a two story, ~3000 sq ft home. I have the Ecobee Suite Manager, Alarm.com Manager, Google Home Community (8 device types), Lock Code Manager (two users), Rule Machine (eight, very basic rules), and Simple Automation (14 rules). The Alarm.com app is set to 9 minute refresh intervals and Ecobee is set to 30 minutes.
I initially had Ecobee and Alarm.com set to lower intervals and was getting the CPU High Load alert about every 12 hours. Even since reducing their polling frequency, I'm still getting the alert every other day. I know when I get the alert fairly quickly as my ZigBee motion sensors stop working. The app and device stats pages don't show anything alarming.
So now I'm wondering, am I just overloading the hub and need to add a secondary?
Particularly with things that poll it can be easy to overload the hub. The runtime stats may point to what is likely causing just be careful with the info as it may not relate to cpu usage easily. Have you checked your hub load with the hub information driver?
I have overloaded my C7 with short polling intervals with the ping driver and the roku connect app.
No, but I'll load that up.
You wouldn't have happened to jump to ADC for security did ya? I used your ADT Tools app on SmartThings.
No for security I took the all in on hubitat approach and it is handled by Hubitat Security Monitor. A purists would argue it isn't the best security approach since it is running on home automation system, but at this time I am not looking for professional monitoring.
With that said every part of the security aspect except the siren is handled by Ring Gen 2 security devices. Simply put I have Ring Gen 2 keypads, Ring gen 2 motion sensors, and Ring gen 2 contact sensors to replace all the dual branded ADT Smartthings gear. If tomorrow I wanted it all to have professional monitoring i could obtain and hook up a basestation/Ring siren and then it would be a full ring alarm.
So far it has all been mostly perfect with a few obvious limitations.
I also use Node Red with the Hubitat and Samsung Automation Studio pallets to integrate between samsung and Hubitat. This is largely used for Arlo cameras i have. It was a bit to setup because of allowing Smartthings events to hit a local endpoint, but has been show to work pretty flawlessly.
The Arlos will even trigger their sirens if the Alarm goes off. So actually have 4 sirens in total.
2 Likes
I actually have a thread discussing this topic as well with a second hub for heavy LAN connected devices. Zigbee and Zwave devices seem to have little impact as long as they are talking to the hub right(ie zwave ghost devices not being present).
I picked up a C4, and have been experimenting with with it. It has honestly been a mixed bag as the complexity has of managing everything does get a bit more complicated with Hub mesh or whatever you use to make them work together.
The saving may also not be what you expect if you get the polling right for the Lan devices. I saw a very small decrease with load once the polling was set to a good level.
My heavy impact items i found were the Hub Ping Driver, Roku Connect app and devices, and the Ecobesuite and i have 5 helpers with the Ecobee suite app. The chromecast integration can also use a bit of cpu for the google home devices. I only have 3 vs your 8 so that could add up as well. Two of the helpers in particular seem to be very heavy, and they are based on the Smart Circulation Helper app. Action Tiles also seems to be kind of heavy,
After my experience on the C4 I wouldn't be so fast to jump onto a second hub. I would get the hub information driver loaded so you can see the load maybe even get something like hubigraphs so you can visualize it over time.
The biggest take away from the C4 i have is that the more devices you can keep on zigbee/Zwave instead of lan the better.
1 Like
I was just curious about the alarm side. I went with ADC through Surety to get professional monitoring that included fire. There's an ADC integration app similar to what you did for ADT, but ADC changed something on their side that's causing intermittent authentication issues when interfacing with the website. I found that you can interface with the Qolsys 2 panel directly across the LAN through a websocket, but they enforce TLS with a self-signed, expired certificate. I haven't found a way to get Hubitat to ignore the cert validation.
Back to the CPU, Ecobee Suite was a heavy hitter when I first installed it. I started getting the CPU load shortly after. I only have two helpers setup and all they do is change the thermostats mode in the morning and at night. I removed all the remote sensors from the app and cranked polling up as high as it would go.
I did get Hub Info setup with Hubigraphs last night and woke to the CPU load alert this morning. Based on the graph, it almost looks like a process is getting stuck. Green is CPU and blue is 5minCPU.
Just for reference with my Ecobee suite install I still have the polling at default and I use 3 sensors. Though it is a bit of a hog it isn't doing anything like causing the cpu to spike like that. To me this looks like it may be a race condition occurring. Considering that Ecobee can function completely without Hubitat have you tried uninstalling it to see if this still occurs.
I would also suggest you take a look at the runtime stats to see what are the big contenders for these events. Now that you have these graphs to it might be nice to try to zero in on the time intervals where this occurs. If the time the event occurs is predictable maybe clear the stats right before you expect the event to occur and then watch the stats for what appears to be the big hitter.
It would be interesting to compare day over day to see if the pattern matches. looks like around a 3.5 hour gap between the spikes.
I noticed the interval as well. I'm starting by removing ESM and doing a reboot. Last reboot was the evening of the 25th, so I'll just have to keep and eye on it for a few days.
Four days of uptime with no CPU load issues after removing Ecobee Suite Manager. I'm going to re-enable it tonight and monitor again. We'll see what happens.
So much for that. Didn't even make it 24 hours. I re-added ESM, changed the polling to 30 min, only added the two thermostats, and setup two mode change helpers that watch some virtual switches. The spikes correlate to polling events.
So it sounds like ESM is causing the problem. Did you report that to the developer of this app? And I would suppose you are running the latest version of that app?
Just scanning the forums, it looks like there are many reports of ESM causing (or suspected of causing) slowdowns going way back.
I don't use this app, so I can't offer any specific advice beyond reporting this to the developer in the appropriate thread for this particular app.
3 Likes
What happens if you just install ESM and let it setup the thermostat and sensors, no helpers involved.
Maybe also do just one thermostat at first.
Berry is pretty responsive to issues. I would certainly post it in the main thread.
ESM can certainly appear to be a hog because of the runtime times involved with the cloud. Calls to the ecobee cloud are not the fastest.
I just posted over in the ESM thread, so we'll see what the dev there says.
