poor man`s cluster in a box with raid1 across nfs

i have some idea in mind i´d like to share before starting a test setup (sort of a "cluster in a box" or "shared nohing san")

scenario:

- 2 esx boxes with local storage only

- each of those boxes hosting a linux vm which exports the local storage as a nfs datastore which is mounted on both esx servers as.

now create a vm which has it`s two identical system disks placed on nfsdatastore1 and nfsdatastore2 and create a software raid1 inside those vm (linux mdraid e.g).

so, each write inside that vm should go to both esx servers, travelling trough both "linux nfs vsa`s".

if i create a copy of the vm`s configuration files (so it exists on both nodes,) in theory i should be able to manually failover a vm to the "surviving" node.

I`m curious if i should spend time on giving this a  try or if there is some gotcha i missed - so - what do you think?

would this work?