I created a script to manage the glusterFS volume-claims:

if [[ "$#" -le 3 ]]; then
  echo "Usage:"
  echo "   $0 <operation> <namespace> <name> <size>"
  echo "   - operation: create | delete"
  exit
fi

OPERATION=$1
NAMESPACE=$2
NAME=$3
SIZE=$4

function create {
  gluster volume create $NAMESPACE-$NAME replica 3  server1:/mnt/gluster-storage/brick-$NAMESPACE-$NAME server2:/mnt/gluster-storage/brick-$NAMESPACE-$NAME server3:/mnt/gluster-storage/brick-$NAMESPACE-$NAME  
  gluster volume start $NAMESPACE-$NAME

  kubectl -n $NAMESPACE apply -f /etc/kubernetes/glusterfs-endpoints.yml
  kubectl -n $NAMESPACE apply -f /etc/kubernetes/glusterfs-service.yml

  cat <<EOF | kubectl apply -f -
apiVersion: v1
kind: PersistentVolume
metadata:
  name: $NAME
  namespace: $NAMESPACE
spec:
  capacity:
    storage: $SIZE
  accessModes:
    - ReadWriteMany
  glusterfs:
    endpoints: glusterfs-cluster
    path: $NAMESPACE-$NAME
    readOnly: false
  persistentVolumeReclaimPolicy: Retain
  claimRef:
    namespace: $NAMESPACE
    name: $NAME
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: $NAME
  namespace: $NAMESPACE
spec:
  accessModes:
  - ReadWriteMany
  resources:
     requests:
       storage: $SIZE
EOF
}

function delete {
  kubectl -n $NAMESPACE delete pvc $NAME
  kubectl delete pv $NAME

  yes | gluster volume stop $NAMESPACE-$NAME
  echo
  yes | gluster volume delete $NAMESPACE-$NAME
  echo
  echo "#################################################################"
  echo "REMOVE BRICKS MANUALLY:"
  echo "  server1:/mnt/gluster-storage/brick-$NAMESPACE-$NAME"
  echo "  server2:/mnt/gluster-storage/brick-$NAMESPACE-$NAME"
  echo "  server3:/mnt/gluster-storage/brick-$NAMESPACE-$NAME"
  echo "#################################################################"
}

case $OPERATION in
  create)
    create
    ;;
  delete)
    delete
    ;;
esac

This creates the gluster volume and maps it to a volume claim in Kubernetes. So you can use glusterfs mounts without the need for automatic provisioning.

Make sure to run kubelet outside docker otherwise you'll be using an old version of gluster.fuse that misses a lot of the optimizations of the modern versions.