That's right PDL / APD state relates to a loss of storage connectivity - e.g. access to LUN has been removed from the host. In that case you can try to restart the VMs on a non-affected host.
For PDL the host has received a message from the storage system saying the LUN is no longer available. For APD this has not happened and so is more of an unknown state and as such you can configure different responses for each event including a timeout for APD to give the situation some time to rectify itself in case it is a transient issue.
Thank you John,
If the host cannot talk with the storage becuse PDL. How can the vm will restart on other host if the VM lost thier VMDK which is been located on the affected storage ?
It's not that the VMDK is lost, more unavailable or inaccessible. So for example you have a cluster with 2 hosts. Someone accidentally modifies the zoning so that LUN-A is no longer presented to Host-1. In this case this would be flagged as PDL and the affected VM or VMs (which are hosted on LUN-A) would be restarted on Host-2 which still has full access to LUN-A.
Or someone accidentally pulls out the cable from the HBA this would result in the loss of accessibility to the storage array from that host and VMs would be restarted on another unaffected host which still has full access to the storage.
It's the same principle as VM restarts with standard HA but standard HA wouldn't pick up these events as the host is still up and working and accessible on the network.