What's the meaning of : from here: https://pve.proxmox.com/wiki/Storage:_GlusterFS ? that's absolutely wrong. Gluster doesn't need to do a full sync (and also it doesn't use rsync at all except for geo-replication) after a brick failure. Only missing files from that bricks are synced back, like in every other replicated system. Even in a RAID, if you loose a disk, you have to sync the whole disk. Additionally, Gluster with sharding is absolutely suggested for VM images. Sharding is developed exactly for VM images, as it decrease the healing time a lot and avoid full image locking.