Reference:
The primitive role roles/viewer provides read access to all resources in the project. The permissions in this role are limited to Get and list access for all resources. As we have an out of the box role that exactly fits our requirement, we should use this.
Ref: https://cloud.google.com/resource-manager/docs/access-control-proj It is advisable to use the existing GCP provided roles over creating custom roles with similar permissions as this becomes a maintenance overhead. If GCP modifies how permissions are handled or adds/removes permissions, the default GCP provided roles are automatically updated by Google whereas if they were custom roles, the responsibility is with us and this adds to the operational overhead and needs to be avoided.

From this source: https://cloud.google.com/billing/docs/how-to/custom-roles#permission_association_and_inheritance
"For example, associating a project with a billing account requires the billing.resourceAssociations.create permission on the billing account and also the resourcemanager.projects.createBillingAssignment permission on the project. This is because project permissions are required for actions where project owners control access, while billing account permissions are required for actions where billing account administrators control access. When both should be involved, both permissions are necessary."

https://cloud.google.com/compute/docs/instance-groups/autohealing-instances-in-migs#setting_up_an_autohealing_policy

This is the most optimal solution. Rather than recreating all nodes, you create a new node pool with GPU enabled. You then modify the pod specification to target particular GPU types by adding node selector to your workloads Pod specification. YOu still have a single cluster so you pay Kubernetes cluster management fee for just one cluster thus minimizing the cost. Ref: https://cloud.google.com/kubernetes-engine/docs/how-to/gpus Ref: https://cloud.google.com/kubernetes-engine/pricing Example:
apiVersion: v1
kind: Pod
metadata:
name: my-gpu-pod
spec:
containers:
name: my-gpu-container
image: nvidia/cuda:10.0-runtime-ubuntu18.04
command: [/bin/bash]
resources:
limits:
nvidia.com/gpu: 2
nodeSelector:
cloud.google.com/gke-accelerator: nvidia-tesla-k80 # or nvidia-tesla-p100 or nvidia-tesla-p4 or nvidia-tesla-v100 or nvidia-tesla-t4

Read the dataset's metadata and to list tables in the dataset. Read data and metadata from the dataset's tables. When applied at the project or organization level, this role can also enumerate all datasets in the project. Additional roles, however, are necessary to allow the running of jobs.
BigQuery Data Viewer
(roles/bigquery.dataViewer)
When applied to a table or view, this role provides permissions to:
Read data and metadata from the table or view.
This role cannot be applied to individual models or routines.
When applied to a dataset, this role provides permissions to:
Read the dataset's metadata and list tables in the dataset.
Read data and metadata from the dataset's tables.
When applied at the project or organization level, this role can also enumerate all datasets in the project. Additional roles, however, are necessary to allow the running of jobs.
Lowest-level resources where you can grant this role:
Table
View
BigQuery Job User
(roles/bigquery.jobUser)
Provides permissions to run jobs, including queries, within the project.
Lowest-level resources where you can grant this role:
Project
to run jobs https://cloud.google.com/bigquery/docs/access-control#bigquery.jobUser databaseUser needs additional role permission to run jobs https://cloud.google.com/spanner/docs/iam#spanner.databaseUser