Patch ESX-1001732 (ESX 3.0.2) and Patch ESX-1002084 (ESX 3.0.1) addresses an issue that can cause the -999 errors:
Fixes an issue where a virtual machine becomes unresponsive on some I/O operations such as taking a snapshot, when the scatter gather element in the I/O thread is not in multiples of 512 bytes.
One possible cause of the above is the use of SQL Server in a virtual machine.. SQL Server, and other database products, can produce high I/O on the virtual disk. The virtual machine may lock up, freeze, or otherwise become unresponsive.
Ensure ESX Server is functioning correctly on all managed servers/Hosts:
Attempt to host a virtual machine through the Virtual Infrastructure Client - a "dummy" virtual machine will work
On each ESX Server, run vmware-cmd -l. This command returns a list of virtual machines on the local server or returns an error if something is not functioning correctly.
If ESX or hostd appears to be malfunctioning, at the service console of the ESX server run the command service mgmt-vmware restart to restart the ESX Server hostd service.
Obtain the vmware*.log files for the virtual machine generating the error by obtaining the VM directory ID from the Properties page of the VMCollect and analyze the vmware*.log files found in that directory.
Make sure that there is enough free space on the datastore for the operation.
Ensure the vmx file for the virtual machine does not have an entry for workingdir
