When disks are imported through Containerized Data Importer, the data flows on default networks: administration network for VMware and pod network for OpenShift. This can lead to poor performance and network congestion. The goal of this epic is to track the implementation of mechanisms to allow transferring data on specific networks, for both VMware and OpenShift.

On VMware side, the data transfer is always done via the ESXi host. However in the default setup, VDDK connects to vCenter to retrieve a authentication token and then connects to the ESXi host on the address provided in vCenter inventory, i.e. the administration network. So the requirement is to:

• List all the networks available for a given group of hosts.
• Select a transfer network the group of hosts.
• Retrieve credentials for the group of hosts (likely for root).
• Create a new object to store the host configuration: IP address, thumbprint and credentials.
• Translate the Managed Object References between vCenter and the Host, or use names.

On the OpenShift side, the data transfer network must exist in the destination namespace. The NetworkAttachmentDefinition is namespaced, so it has to exist in the namespace where the importer pod is be created, i.e. the destination namespace. When users create a Plan, they specify the destination namespace, so the requirement is to:

• List all the networks in the destination namespace.
• Add the ability to set a transfer network in the VirtualMachineImport CR.
• Extend the Plan CR to add the transfer network to the generated VMImport CRs.