Hello, We have deployed metaflow using terraform. And you are using instances from the following instances family:

c5, r5, g4dn

After deployment when we checked Auto Scaling group that contain

c5

and

r5

instances which got created through AWS Batch, It does not contain all the instances from the mentioned instance family. Does anyone know why is that the case?

I'm not sure exactly in this case, but as a more general comment we don't recommend mixing GPU and non-GPU instance types in the same AWS autoscaling and Batch compute env. In cases like this you want two separate environments and queues, and route tasks to them yourself . See this thread too: https://outerboundsco.slack.com/archives/C02116BBNTU/p1680198775873169?thread_ts=1680191892.868119&cid=C02116BBNTU

It’s not a problem with gpu and non gpu instances. Batch is creating two separate auto scaling groups for that. The main issue is, auto scaling group with cpu instances are allowing only bigger machines like machines bigger than 8xlarge. And the batch desired cpu number is increasing when a new job is starting but it is not reflecting to the auto scaling group.