Background: Working with a customer that wants centralized management as they are providing a type of managed service. They wanted to use global credentials (aws, gcp, azure, global-pullsecret, ssh ..etc) so they placed them all in the hive ns. ClusterDeployments, ManagedClusters, install-configs, klusterletAddOnConfig ..etc are provisioned in the hive namespace.

Problem: The install configs for AWS are using m5.4xlarge but when the clusters are provisioned they are built with m6i.xlarge. I initially thought this was availability in the  region so I tried with multiple machineTypes but kept getting the same thing. Azure and Google environments work and deploy as expected. I tried adding machinepools to see if this would change the behavior and although i could scale to the new machinetypes I could not provision with them.

I moved the resources to the cluster.name namespace and everything works as I would expect. Machines are provisioned with the right machineType.

Is this behavior expected?