Back in the days of ESX 2.5 I managed a bunch of IBM Servers. It regularly happened that when a raid mirror set broke the IBM ESX host(the only servers I witnessed this with…) froze completely. I thought this problem was fixed with 3.0.x and I thought it was a VMware problem. Yesterday one of my customers had the exact same issue with an IBM server and ESX 3.0.2. The host froze completely and the state was “disconnected” while leaving the VM’s online. It was caused by a defect harddisk, which broke the mirror. It wasn’t possible to reconnect to the server. I guess it’s due to the firmware of the scsi/raid controller, and this topic on the forum probably confirms my suspicions. Let’s see if I can reproduce this and solve it by flashing the raid controller.