Description of problem: Ideally gluster would disable online discards (mount -o discard) but would periodically discard the individual bricks' mount points via fstrim(8). Online discards are slow and generally undesirable but there are benefits to the underlying storage if the address space is periodically discarded (unused space is reclaimed). If/when gluster makes use of DM thin provisioning and DM cache, periodic discards (during scheduled downtime/off hours?) will enable the underlying storage layers to be more performant. (I have no idea if I picked the right component for this BZ, but I discussed the general RFE with Brian Foster and did my best to describe what we'd like to see.. please feel free to change component as needed).