Under certain circumstances some or all VMs in a vSphere environment may show the red alert icon and an alert status in a VM list, but no alarm in the VM properties that could be acknowledged or cleared. These “phantom alerts” are an issue for quite some time now and still not fixed.
They seem to be caused by an interruption of the connection to the datastores these VMs reside on, for example by an ESXi host booting while the storage is not up & running. The VMs are first displayed as “disconnected”, but after the datastore connection is restored, the alert status stays.
There is no straightforward way to resolve this, but a vMotion clears the status, most likely because the VM registration is updated. So basically just move the VMs around to clear the alerts, the quickest way is to set the hosts into maintenance mode and evacuate them. A host reboot is not required.
There seems to be a serious problem in the update process of the vShield Manager appliance. I had to update a vSphere 5.0 / vCloud Director 1.5 environment running vSM 5.0.1-638924 to the latest releases (vSphere 5.1 update 1a / vCD 5.1.2 / vSM 5.1.2a). Disk space on the vSM was sufficient, the update itself using the VMware-vShield-Manager-upgrade-bundle-5.1.2-943471.tar.gz went fine, the appliance rebooted and was busy with the upgrade process for some time, as expected. Afterwards the UI showed the new release, the VM was idling, update apparently successful.
But after a manual reboot (due to a scheduled maintenance of the whole environment) before migrating the configuration to a new vSM instance the appliance would not come up anymore. The disk driver encountered fatal I/O errors, maybe due to file system or partitioning damage (although the symptoms do not clearly indicate that), finally resulting in a kernel panic: Continue reading