Hi,
First, I would like to describe shortly the issue:
Storage team removed LUN, but datastore was still mounted to the ESXi hosts within cluster. Hosts hadn't rescanned HBA after LUN removal. This caused repeated situations where all ESX's that had this datastore mounted became unresponsive with 100% CPU usage affecting hosted VM's and overal operational health. Rescan storage on each ESX solved problem. Please note that datastore was in maintenance mode and empty , all the time from the start. ESX version is 7.0.3n
Question: Why it happened impacting production VM's ? This is normal or maybe known ESXi issue ? CPU starvation or ???