Yes it requres a shared disk i.e. isci it also requires block level access there are further requirements i.e.
a failed node must not become available while another node is recovering the journal for the failed node https://en.wikipedia.org/wiki/GFS2
TomerLeibovich
ASKER
What about gluster?
Does it also requires shared disk?
Highly available always requires something being shared. This way if your compute node goes down then the remaining compute node still has access to the data.
Gerwin Jansen
Where would you run the (2) Linux VM's? If you run them on the same host and the host fails then you have no storage (you could use local storage on the host as shared storage). If you run the VM's on separate hosts then you need a third host with the (shared) storage. How do you intend to spend the budget that you have? You do have a budget right?