I have some issues with my AKS setup, I have some pods running Openresty (nginx) where they are receiving huge amount of requests. Autoscaling works fine however I am experiencing issues with one case - when node pool has more than one pod like 2 or 3 everything is working as expected however when there is a peak and AKS adds a new node pool to create additional pod (single node pool with one pod running) Load balancer during round robin distribution sends the traffic to a new node pool making this pod over-utilized and causing memory limits.
During further autoscaling additional pod is added to the node pool and problem does not exsists anymore however this corner case is a bit frustrating
Here is Grafana screenshot
