1

We've determined a base metric that the servers in our ASG can handle about 1000 requests per minute.

Given that information, we'd like to be able to control the scaling of our ASG based on the RequestCount metric on the ELB. We'd like to do something like:

Number of servers to run = (RequestCount / 1000) + Some extra servers for "padding"

Basically see how many servers we must run given our estimation on requests per minute that each can run, then maybe add a few more to handle smaller spikes better.

However, I am not seeing how to do this via the AWS scaling rules. It seems like we'd have to make an alarm, but since we want to dynamically scale based on the value, there isn't really a threshold we can set for when the alarm should "go off".

Is there a way to do this kind of scaling out-of-the-box with the core AWS tools and UIs, or would we need some other kind of tool or service to achieve this?

jzimmerman2011
  • 323
  • 1
  • 4
  • 12

1 Answers1

0

You can create a custom cloud watch metric to monitor the connections, Create an Alarm based on that custom metric, then based your autoscaling on that alarm.

Creating custom cloudwatch metrics

Or you can use the ELB ActiveConnectionCount metric in cloudwatch, create alarm, edit the scaling policy and pick that metric alarm.

https://docs.aws.amazon.com/autoscaling/latest/userguide/policy-updating-console.html

https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/elb-metricscollected.html

strongjz
  • 832
  • 4
  • 7