Prepending important information to the well answered original post below:
IPMI requires hardware access that cannot be logically hypervised.
Any IPMI interaction you would be doing would be the client on the guest OS connecting to it's available physical hardware - which is
really a 'service' to a device that is listening to connections while
the ipmidev is loaded in the kernel.
Since the BIOS itself is virtualized by the hypervisor, there is no physical access directly to hardware on the host OS. Thusly, since
there is no BIOS to access hardware directly, serial redirection to
hardware cannot be possible.
I'm not aware of any hypervisor which would have implemented IPMI interface. So this may be the reason why you can't start the OpenIPMI service as it tries to load IPMI kernel modules which can't obviously detect it.
In my opinion, you don't need to have this service running in your management virtual station when you want to manage servers over network from it. You just need to have enabled and properly configured IPMI interface on service processors of your servers which are iDRAC for Dell servers, iLO2/3 for HP servers, ALOM for Oracle and so on, installed OpenIPMI tools on that station and then you should be able to connect to service processors and talk to them with IPMI protocol (most of them are compliant with IPMI standard). Below is an example how to query a server for chassis status:
ipmitool -I lan -U root -H 192.168.1.1 chassis status
Or you can power off the server:
ipmitool -I lan -U root -H 192.168.1.1 -a power off