When the VMs failover, then the VMDKs become visible and we can protect those VMs, before then as Ali stated, the VM disks are invisible to the VM and to us. The clumsy way to protect them automatically is to create a separate subclient and allow the job to run every night. It will fail, but when the failover occurs it will succeed. You can disable the subclient from SLA to avoid it impacting your reporting - but its not the best solution.
Since SRM is all about recovery automation, the proper solution would be to have SRM enable the schedule to backup the failover site as part of its recovery plan. You could script this using our command line tools. Here are some examples on how to enable/disable a schedule:
https://documentation.commvault.com/11.24/expert/6712_managing_schedules_and_schedule_policies_using_command_line_interface.html