Remove img_handlers config parameter usage

Now that the only disk code requiring host filesystem
mounts is the LXC filesystem setup, the img_handlers
config parameter is an solution looking for a purpose.
For any disk image format we can categorically say
whether it should be mounted using loop or qemu-nbd
mount impls. There is no desire to use libguestfs FUSE
for mounting LXC disk images, since users of LXC do
not want a huge KVM process alongside every container!
Move the mount class choice logic into the mount API,
by creating new methods

  nova.virt.disk.mount.API.instance_for_format()
  nova.virt.disk.mount.API.instance_for_device()

and then call these from nova.virt.disk.API code

blueprint: virt-disk-api-refactoring
Change-Id: I088b5debdf62d4a9a6b4521eee68cd4757f2ff42
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
This commit is contained in:
Daniel P. Berrange
2012-11-13 11:15:40 +00:00
parent a1a5ebbbd6
commit f7a17b9af7

View File

@@ -107,7 +107,6 @@ class TestVirtDisk(test.TestCase):
'/mnt/loop/part': '/dev/mapper/loop0p1',
'/mnt/nbd/nopart': '/dev/nbd15',
'/mnt/nbd/part': '/dev/mapper/nbd15p1',
'/mnt/guestfs': 'guestmount',
}
return mount_points[mount_point]
@@ -141,11 +140,4 @@ class TestVirtDisk(test.TestCase):
('qemu-nbd', '-d', '/dev/nbd15'),
]
disk_api.destroy_container('/mnt/guestfs')
expected_commands += [
('fusermount', '-u', '/mnt/guestfs'),
]
# It's not worth trying to match the last timeout command
self.executes.pop()
self.assertEqual(self.executes, expected_commands)