Description of problem:
Using autofs to mount gluster volumes, the backupvolfile=$backup-server option is ineffective, though this works correctly when calling mount -t glusterfs directly from the command line. Specifically, given an automount entry in as:

volume  -fstype=glusterfs,backupvolfile-server=server2 server1:volume

If server1 is down, volume will not mount. However, issuing

mount -t glusterfs -o backupvolfile-server=server2 server1:volume /mnt/volume

DOES work. For a specific case, here are relevant lines from the glusterfs log file (the actual mount point is /misc/shared in this case):

[2013-07-27 00:25:44.848810] E [socket.c:2157:socket_connect_finish] 0-glusterfs: connection to 192.168.10.228:24007 failed (Connection refused)
[2013-07-27 00:25:44.848893] E [glusterfsd-mgmt.c:1851:mgmt_rpc_notify] 0-glusterfsd-mgmt: failed to connect with remote-host: Transport endpoint is not connected
[2013-07-27 00:25:44.857882] E [fuse-bridge.c:4973:init] 0-fuse: Mountpoint /misc/shared seems to have a stale mount, run 'umount /misc/shared' and try again.
[2013-07-27 00:25:44.857934] E [xlator.c:390:xlator_init] 0-fuse: Initialization of volume 'fuse' failed, review your volfile again

Version-Release number of selected component (if applicable):

This is happening with gluster v 3.4.0-3, installed into rhel6.4 64-bit, as well as on Fedora 18 clients. All of the servers are rhel6.4 x86_64.
rpm -qa|grep gluster reports on F18:

glusterfs-server-3.4.0-3.fc18.x86_64
glusterfs-3.4.0-3.fc18.x86_64
glusterfs-fuse-3.4.0-3.fc18.x86_64


How reproducible:
1. Create a replicated gluster volume.
2. Set up an automount map with a primary server and backupvolfile-server pointing to an alternate.
3. Take the primary gluster server off line
4. Attempt to mount the volume on a client via automounter.


Actual results:
Volume does not mount.

Expected results:
Volume should be mounted from the backup server.


Additional info:
The logfile notation of "Mountpoint /misc/shared seems to have a stale mount, run 'umount /misc/shared'" was suspicious, so I patched /sbin/mount.glusterfs to unmount before trying the backup server:
--- mount.glusterfs~    2013-07-26 20:25:34.028813360 -0400
+++ mount.glusterfs     2013-07-26 20:26:45.849772714 -0400
@@ -227,7 +227,7 @@
         err=1;
         if [ -n "$cmd_line1" ]; then
             cmd_line1=$(echo "$cmd_line1 $mount_point");
+            umount $mount_point
             $cmd_line1;
             err=0;

and with this, automounting from the backup server works ~80% of the time.