We have a cluster built via Rancher, I have monitoring enabled and I can see the monitoring showing exactly the double of used memory as the "kubectl top pod" command, and this is causing memory used higher than limit and impossible to deploy an HorizontalPodAutoScaler on memory usage...

Anyone has ever been in this problem or knows where should I star looking for the problem?

# kubectl top pod xxxxxxx-api-66f8446df9-drccw
NAME                             CPU(cores)   MEMORY(bytes)   
xxxxxxx-api-66f8446df9-drccw   113m         310Mi  

And Rancher UI:

Rancher     v2.3.5 
User Interface  v2.3.36 
Helm    v2.14.3-rancher1 
Machine v0.15.0-rancher29