Cluster load balancers are not spinning up because they are having trouble finding suitable subnets. This is most often seen in in HCP with a degraded console cluster operator. In ROSA classic, we don't see this as often, but the problem of load balancers not being able to find a suitable subnet can still happen.

The subnets are tagged kubernetes.io/cluster/<Cluster Identifier> for _another_ cluster or using the cluster's name. The subnet selection process prefers clusters tagged for the current cluster and excludes subnets tagged for other clusters.

https://github.com/openshift/cloud-provider-aws/blob/3afccd99e97fdab41c36996575ea87f6d48ab389/pkg/providers/v1/aws.go#L3490-L3524

    1. Tag subnets `kubernetes.io/cluster/asdfasdfasdf`
    2. No suitable subnets will be found for load balancers
    3.
    

Solutions:

1. Remove all kubernetes.io/cluster/.* tags. This will allow the subnets to be considered equally. This may not be possible for all BYOVPC customers as they may have a subset of subnets they want to use for clusters.

2. For both HCP and Classic, the subnet must be tagged with the cluster's "infra id". For classic that is ocm get cluster $CLUSTER_ID | jq '.infra_id'. For HCP, that is the internal ID.