I was looking into the various graphing options to show how memory leaks were affecting my main hub. All are kind of a lot of work and I realized all I really wanted to know was "how big is the leak now and how long until my hub will be unstable?" I've done a lot of work to reduce the memory load in my main hub and this allows me to see that things are going much better. Sharing the excel workbook in case anyone else is interested.
I'd like to expand on this topic. Does anyone have a way to measure the memory impact of certain apps and/or devices? I think the more useful thing is tracking down what's leaking memory so that we can either get it fixed or remove it if its not essential. I'd love to be able to log memory usage on a per-app basis.
My hub used to leak memory and required a weekly reboot. Then I changed a bunch of things and it no longer leaks memory and will run reliably till its next software update. Unfortunately, I changed a lot of things at the same time and I have no idea which one removed the memory leak.
Those are the only apps I use. I’m rebooting every 2 or 3 days. It turns into a week+ if I disable z-wave polling. It’s polling two dimmers that don’t reliably report on/off/level state otherwise.
I have tried rate limiting the accesses to the Maker API endpoints. My good night macro sends about 30 requests to it.
I’m not going to buy an upgrade from my c5 since it’s not faster or more RAM. I’ve taken to implementing anything that would be a plugin for the Hubitat as stand alone microservices running in Docker containers because the risk of shortening the uptime between rebooting is way too high.
For example, I uninstalled the sleep number plugin and wrote my own, and that extended the uptime by a day.