Insights scales

I’m looking into some insights and am a bit confused by the scale it works with. Looking at average system load it goed from 0 to ± 2.5. So is for example “1” equal to 1%, or 10%, or is 1 “full load” and everything above overload? The last would make the most sense in my case.

Is there Anyway to format the scales within the insights pages (so no exported data)

You are correct: 1 means “full load”, and everything above it means your system is overloaded. A load of 2.5 means that (on average) 2.5 processes are waiting to be assigned a CPU time slice.

It’s a bit confusing that in Insights, the “load average” value is used (where 1 = full load), but in the mobile Homey app, a percentage is used (which, AFAIK, is basically the load value * 100).

I don’t think there is.

1 Like

Quite odd that there are so much spikes during the night, but the part between 0800 & 0900 explaines why things don’t go smooth, or not happening at all.

Time to try and shed some more weight while athom makes a backup and restore function so I can move to a homey pro or even better a homey 2.

Tnx

I also see a fair amount of spikes, I think that Homey’s core periodically does housekeeping at those times.

You could try to correlate those spikes to specific apps (although, again, it’s confusing that for apps, CPU usage graphs are incremental, meaning that instead of looking for peaks, you need to look for sharp increases).

Yeah have been looking but there is no real correlation te be found. The spike of above 2 can be explained. A door was opened while the alarm was armed and then a bunch of things should happen like turning on all lights (different manufactuors) , wich have always been a bit heavy for homey. Now however it was more miss then hit, wich is the second time now.

I’ll make this a bit lighter by removing the fancy parts wich are not strictly neccesary.

Also, load average != CPU usage. Load average is the pressure on the scheduler, where 1 means it can execute as much as is being scheduled on one core. CPU usage is how much of a cores time is spent in each state. A single core machine can’t have a CPU usage higher than 100%, but it can have a load average much higher than that. I am not sure if load average ~ CPU usage for values much lower than the number of cores, but it is definitely not so for values close to or above the number of cores.

1 Like

Both the percentage values and the values shown in the Insights graphs are based on the load average (where using % as a unit makes no sense as I’ve tried to explain to Athom once, but they don’t particularly care).

The Insight graphs for apps are even more confusing, because they use % as a unit but are based on some sort of incremental metric (the graphs always go up), which, I assume, it based on CPU time. However, how that value is translated to a percentage, I don’t know.

1 Like

Same subject, but diff. angle of scaling…

Is there any way I could impact insight scale (only issue with web, android app OK) - e.g. here 9.0 when clearly off (actual trend value have 4 decimals)

image

Trend values from AVD / number fields if makes difference somehow…

Just in case some bothered with same issue (to close the loop): When creating trend, take the 1st items with desired desimals defined. Seems that scale always based on 1st one (only)…