Adjust memory "slop" associated with a vnode host.
We hit a scenario where we were using 126GB of RAM on a shared d820 vnode host (117GB for vnodes, 8GB for dom0 and 1GB "slop") leaving 2GB for additional vnodes. Someone tried to start 2 x 1GB vnodes, which both mapped to that host, but the second one failed to start with
xl create complaining that it ran out of memory for the domU's "reserved memory". So perhaps 1GB is not enough "slop"? Given that
d710s only have 12GB of physical RAM, lopping off another 0.5 to 1GB is a significant hit. But maybe we don't care about
d710s as vnode hosts anymore.