This explains why my C-3's Hub Watchdog-measured VS delay roughly doubled when I updated from 2.2.5 to 2.2.7 (I skipped a release). I had been running 2.2.5 with all event and state history sizes pared back to low single digit numbers. After the 2.2.7 update, I was surprised to see for the first time in recent history the hub's performance objectively decrease with a new firmware... and after a few days uptime, I noticed System Events indicating that the Zigbee radio was periodically cycling off and on (roughly 3 minutes on, 6 minutes off, repeatedly) accompanied by 'Severe CPU Load'.
Soft reset didn't improve anything... glad I happened to read your post; otherwise I'm not sure how (other than regressing to 2.2.5) I would have been able to fix this. Definitely merits some kind of publicity...
The fastest my C-3 could flip a virtual switch on 2.2.7 was .059 with the truncated history sizes; after bumping them back up, it can now match the previous min of .028 (my ancient C-3 with roughly 80 real devices and WebCoRE with >100 pistons runs quite fast... for some reason, even my sandbox C-7 with nothing on it but Hub Watchdog can't match its performance).
If the increased GC activity had bought me any more uptime it would be worth it (aside from the whole Zigbee radio snafu), but the hub still needed to be rebooted roughly every 5 days before measured delays began to increase (that's about the same 'good performance' uptime I'd had on 2.2.5).
My issue with trying the troubleshooting is it takes days for my hub to get to severe load so if I disable an app to see if it is causing it I can not use that device or app for DAYS!!! I can not go without webcore or ecobee etc.
That is part of what makes it tough. Although on the flip side, if it takes days before it becomes a problem, just setup a rule and reboot nightly. I did that for quite a long time until a few kinks got worked out in my logic (and in the hub software).
@JasonJoel I think I have figured it out. I have removed the ecobee suite and for over a week I have not received the severe load message. But now that i figured that out how do I use my ecobee with hubitat and not get this message.
So I would check on that HA thread I linked to for their thoughts on doing this.. the short answer is yes or in Docker. Looking at it again I don't see a thermostat listed for the HA integration but there may be one added now dunno.
I have a simple HA test set up running on an RPi3..
edit: It's very strange that a supposedly "open" system like Ecobee wouldn't have more integrations..
Have you tried using the built-in Ecobee integration? I have been using it for quite some time and do not have any issues with my hub. It is not as full featured as the Ecobee Suite, though.
It might be a good idea to post this on the Echobee Suite developer's thread. There have been a couple others with similar issues lately. Maybe this is something the developer can fix, or could at least look into?
Well, that's not exactly true. I have seen your reports of concerns about Ecobee Suite (scattered over 3 or 4 different threads, making it difficult to track the context, BTW). From what you've shown me, the CPU and Memory overhead of Ecobee Suite is infinitesimal, even while it may be the busiest application. In fact, MANY ES users run with higher load than you are seeing, and they haven't reported similar issues.
BTW: Shooting the biggest thing you see may have made the problem temporarily go away, but you haven't necessarily found the culprit. Hopefully Hubitat support can help you track down why you are seeing things get clogged up...
If you'd like some of my help regarding what you are seeing, please post the following over in the current Ecobee Suite thread, and I'll see what I can do:
Versions of Ecobee Suite Manager, Thermostat and Sensors
The number of Ecobee thermostats and sensors you are using
Your current polling frequency (set in ES Manager)
The list of ES Helper applications you are running
The screen shots showing the run times for both ES Manager (application) and ES Thermostat/Sensors (devices)
The type of Hubitat hub you are running these all on (C4, C5, C7, ...)
A Snapshot of your Live Logging for ES Manager and your ES Thermostat(s)
FWIW, you don't have to use my Ecobee Suite if you don't like it. You can choose to use the native Hubitat support, or you can pay to use the other user-contributed Ecobee Device setup. I don't make money on Ecobee Suite (it's a hobby), but it meets the needs of quite a lot of people.
And complaining in multiple, unconnected threads makes it extremely difficult for anyone to follow and/or help you.
So Hubitat support told me to remove apps until it goes away. So that is what I did. Email is below. I did everything they said in that order.
Hello there,
Thank you for your time and continued support. If you didn't do so already, please unplug the power cord for 30 seconds before plugging it back in. If that doesn't resolve your problem, then I suggest performing a "Soft Reset" by following instructions in below document:
If the problem persists, and you have any custom code installed on your hub, you may want to reboot the hub in "Safe Mode," from the "Diagnostic Tool," then disable your apps or drivers to identify which causes the problem.
To access the Diagnostic Tool, please follow instructions in below document:
I only have 1 ecobee thermostat. I removed the sensor. I have the most up to date ecobee suite software. I just readded it today to see if it comes back. Now this time I only installed the ecobee suite and none of the add ons.
I have a hubitat C7 Hub.
As for putting it in 2 threads I started it in here and just moved to the other thread.
Here are the live logs but I just put the ecobee back on hubitat this morning.
These are the runtime logs - check your Live Logging to see if anything is generating errors…error handling can really mess up memory on these little machines.
After much research, thanks to @FriedCheese2006, we have determined that the severe CPU load increase is related to lengthy ES Manager cycle times/polling cycles. Tentatively, the solution is to set your polling cycle to 1 minute.
I will research as to why this is happening, but until then my recommendation is to use 1 minute polling cycles whenever using Helpers and/or automations with Ecobee Suite.
