While the attack is running, switch back to Container Insights. If you don’t see a CPU spike after a short time, try reducing the time period of the chart. While viewing the metrics, ask yourself:
We can see on the Container Insights’ CPU Utilization graph that the CPU spike was seen, but the graph for Number of Nodes saw no increase.
Does this mean our experiment failed? Not at all! By running this experiment, we found a potential failure mode before it had a chance to impact customers. We now have an action item to reconfigure our autoscaling rules. We can then verify them by repeating this experiment.