What is the correct way to reduce the size of an Auto Scaling Group in AWS?

Question

Amazon Web Services

Hello!
I have such a situation. There is an Auto Scaling Group, the size of which is formed depending on the number of messages in a particular queue. It is not difficult to write a rule for increasing this group - we see the accumulated messages over the past N minutes, we increase it by M instances.
Then I began to think about how to properly shoot instances. It would seem that everything is also simple - the queue size for some time is kept below the threshold value, so everything is fine and we do not need instances. But the trouble is this: when an instance takes a task from the queue, it processes it, and this processing can take quite a long time. Of course, you need to monitor the activity of each instance, but as I understand it, the entire group as a whole is monitored, and when some rule is triggered (say, there is no CPU load), a random instance will shoot back. Please tell me how to do it better or correct me if I'm wrong.
Thank you in advance!

Answer 1

it seems to me easier to raise and kill the necessary instances with a script than to use autoscaling groups for delay job in your case