We were running ECS
1.39.0 and Docker
We recently decided to upgrade ECS to
1.45.0 and Docker
19.03.6-ce and we noticed that our tasks were responding with much less resources allocated based on the task definition (which remained the same before & after the upgrade).
Looking at the screenshot, we can see a noticeable dip in CPU and that is when we rolled out the updated versions (using the same instance type). When we rolled back to our previous AMI, you’ll see the CPU jumps right back to what appears “normal”.
Can anyone help us pinpoint what may be the cause here? We looked at changelogs for both the ecs agent and docker and nothing immediately jumped out. ECS 1.40.0 had an update to optimize cpu usage on the host and I don’t think that would affect our issue here.