why the autoscaling lab of the profesional architect certification stays in 3 instances for around 25 minutes if the desire instances ws 1 and the stress was program for 10 minutes, the first instance should be cooled down and then terminated as seen but this is not shown in the monitoring in service instances
If you remember, the stress program had a timeout of 10 minutes, so from the time an instances starts, it basically has 10 minutes of full CPU load. After that, we have to wait until the average CPU across the fleet comes down low enough to trigger the scale in activity (below 60% average in this lab’s case).
So, with all instances running for 100% for 10 minutes each, we have to have some time at 0% to pull the samples back down. This is a good thing as you want some smoothing of the instances quantity…you don’t want instances going up and down in rapid succession as they try to react to shifting load across minutes.