VMCP is a new technology from vSphere 6.0. VMCP protects storage failures, used typically for block-based storage. Before vSphere 6.0, vSphere HA only protected hosts by monitoring network failures; if the FDM agent is not able to talk with other agents, it can be considered in a fault state by the other agents, and it will be considered in an isolated state itself.

The following two types of storage failure can now be handled:

• PDL: When the storage array has lost access to the device (the device is offline or unavailable) then a specific SCSI sense code is sent to the ESXi server. When PDL situation occurs, the ESXi server will stop sending I/O commands to the device.
• APD: If the ESXi host is not able to reach the storage array at all (no SCSI sense codes are received) then the storage device is marked as unavailable, but since the ESXi server validates the status of the storage array, it will still try to resend I/O commands to the device until APD Timeout is reached. 

When you want to use VMCP, you must configure the Failures and responses tab of the configuration:

The VMCP recovery workflow is shown in the following diagram: