Initial GFS2 support creates a GFS2 file system per compute node that is then shared between the compute node and the rabbit. So if there are 10k computes, you have 10k GFS2 file systems.
This story is a request for creating a GFS2 file system per rabbit per job which is then shared with the computes attached to the rabbit. So if a 10k compute job uses 700 rabbits, there would be 700 GFS2 file systems with groups of one rabbit, ~14 computes accessing one GFS2 file system.
Ported from internal RABSW-858: