I used these as steps in replicating your deployment:

1. Set up the Cloud Storage FUSE CSI driver for GKE: starts with Workload Identity Federation for GKE so that you can set fine grained permissions on how your GKE Pods can access data stored in Cloud Storage.

2. Mount Cloud Storage buckets as persistent volumes: I use this for my reference in proper PersistentVolume and PersistentVolumeClaim which gives correct access modes to your volume mounts.

3. Use your deployment YAML configuration without the added MySQL flags for GCSFuse compatibility and use namespace from step1 as spec.serviceAccountName.

4. Deploying these YAML configurations gives us running status for deployment:

For the answers to your questions:

1. It's due to spec.accessModes of your PVC and PV it should have matched field as ReadWriteMany to have proper attributes.

2. You can force GCSFuse to mount in GKE using the first step.

3. Don't skip the first step to grant IAM roles to the Kubernetes ServiceAccount.