Is Max CPU utilization in aws ecs fargate service of multiple nodes represent one data point i.e. one node's max cpu or sum of max of all nodes over a specified period divided by total nodes. Example Lets say there are 2 nodes in the cluster, over a period of 1 minute, node-1 observed max cpu of 70% and node-2 50%, then max cpu utilization reported for ECS service will be 70+50/2 = 60% or one node's max which is 70% ?
After further reading and inspecting individual json log messages which container sends to cloudwatch to create these metrics, the max cpu utilization is maximum of all data points collected from all nodes for a specified time period i.e. it represents spike on one particular node. Individual nodes send only the CPU Utilized data point and not max cpu utilized.
Sample snippet from individual node insight message.
CpuReserved : 2048, CpuUtilized : 313.5060546875