After a planned full shutdown due to building power maintenance, I had several VMs return with an Alarm icon, but no corresponding alarm detail or visible problem. Initially, the VMs in question all resided on a datastore volume that was inaccessable to the ESXi hosts during boot. This turned out to be that CHAP was enabled on the volume in the SAN but should not have been active. After turning off CHAP and rescanning for volumes from vCenter the VMs stored on that volume came available to start.
But even after starting they continued to show a red alarm icon with no corresponding alarm detail.
We eventually thought to try vMotion. vMotion reported a problem on the validation sequence with the CD drive volume assigned to the VM. In every case, the VMs had an ISO set as their current CD, and the ISO was stored on a folder in the SAN (but NOT the volume that didn't originally connect). We changed the CD to 'use Host device', but the alarm remained. Tried vMotion again, and that worked. And the alarm icon disappeared. vMotioned back to the original preferred host and everything remained normal.
Bug?
Reproduceable issue? (I don't have a lab or test environment, don't ask)
ESXi 5.0 (no updates), vCenter 5 on Windows guest w/SQL Express