1
0
Fork 0
mirror of https://github.com/retspen/webvirtcloud synced 2026-11-09 11:46:49 +00:00
webvirtcloud/vrtManager/tests/test_external_snapshots.py
catborise 3fd422db7e fix(instances): name snapshot files libvirt could not remove
When libvirt cannot remove a snapshot file it merged or discarded, it
only logs a warning and the delete or revert still succeeds. On NFS it
removes them as the qemu user, so an image directory the qemu user may
not write to left every merged overlay on disk without a word.

Revert and delete now compare the files in use before and after and
show a warning naming the ones still on disk. Only a snapshot's own
files (overlays, memory) and overlays on a snapshot point count, and a
disk the revert detached is never named, so a kept disk is not offered
for removal.
2026-10-12 21:41:20 +03:00

499 lines
22 KiB
Python

"""External snapshots go through libvirt's snapshot API; WebVirtCloud only
checks preconditions and picks the flags. libvirt is mocked here; the live
cases are in instances/tests/test_live_snapshots.py."""
import unittest
from datetime import datetime
from unittest.mock import MagicMock
from django.conf import settings
if not settings.configured:
settings.configure(MAC_OUI="52:54:10")
import libvirt
from libvirt import libvirtError
from vrtManager import util
from vrtManager.instance import PERSISTENT_XML, wvmInstance
INTERNAL = libvirt.VIR_DOMAIN_SNAPSHOT_LIST_INTERNAL
EXTERNAL = libvirt.VIR_DOMAIN_SNAPSHOT_LIST_EXTERNAL
DOMAIN = """<domain>
<devices>
<disk type='file' device='disk'><source file='/pool/a.qcow2'/><target dev='vda' bus='virtio'/></disk>
<disk type='file' device='disk'><source file='/pool/later.qcow2'/><target dev='vdb' bus='virtio'/></disk>
<disk type='file' device='cdrom'><target dev='sda' bus='sata'/></disk>
</devices>
</domain>"""
def snapshot_xml(name, memory=False, disks=("vda",)):
devices = "".join(
f"<disk type='file' device='disk'><source file='/pool/{d}.qcow2'/><target dev='{d}'/></disk>" for d in disks
)
mem = "<memory snapshot='external' file='/pool/x.mem'/>" if memory else "<memory snapshot='no'/>"
return (
f"<domainsnapshot><name>{name}</name><description>d &amp; e</description>"
f"<creationTime>1700000000</creationTime>{mem}"
f"<domain><devices>{devices}</devices></domain></domainsnapshot>"
)
def fake_snap(name, memory=False, disks=("vda",), current=False, children=0):
snap = MagicMock()
snap.getName.return_value = name
snap.getXMLDesc.return_value = snapshot_xml(name, memory, disks)
snap.isCurrent.return_value = current
snap.numChildren.return_value = children
return snap
def proxy(status=1, internal=(), external=(), xml=DOMAIN, jobs=None, managed_save=False):
"""A wvmInstance without a connection. internal/external are fake snapshots."""
p = wvmInstance.__new__(wvmInstance)
p._read_cache = None
p._XMLDesc = lambda flags: xml
p.get_status = lambda: status
p.get_uuid = lambda: "11111111-2222-3333-4444-555555555555"
p.get_cur_memory = lambda: 1024
p.refresh_instance_pools = MagicMock()
p._snapshotCreateXML = MagicMock()
p.wvm = MagicMock()
dom = MagicMock()
snaps = {s.getName(): s for s in (*internal, *external)}
def list_names(flag):
return [s.getName() for s in {INTERNAL: internal, EXTERNAL: external}.get(flag, (*internal, *external))]
def list_all(flag):
return list({INTERNAL: internal, EXTERNAL: external}[flag])
dom.snapshotListNames.side_effect = list_names
dom.listAllSnapshots.side_effect = list_all
dom.snapshotLookupByName.side_effect = lambda name, flags: snaps[name]
current = [s for s in snaps.values() if s.isCurrent()]
dom.hasCurrentSnapshot.return_value = bool(current)
dom.snapshotCurrent.return_value = current[0] if current else None
dom.hasManagedSaveImage.return_value = managed_save
dom.isActive.return_value = status in (1, 3)
dom.blockJobInfo.side_effect = lambda dev, flags: (jobs or {}).get(dev, {})
dom.XMLDesc.side_effect = lambda flags: xml
p.instance = dom
return p
class SnapshotListTestCase(unittest.TestCase):
def test_entries_carry_memory_current_and_description(self):
p = proxy(external=[fake_snap("one", memory=True, current=True)])
[entry] = p.get_snapshot(EXTERNAL)
self.assertEqual(entry["name"], "one")
self.assertEqual(entry["description"], "d & e")
self.assertEqual(entry["date"], datetime.fromtimestamp(1700000000))
self.assertTrue(entry["memory"])
self.assertTrue(entry["current"])
def test_detaches_lists_disks_missing_from_the_snapshot(self):
p = proxy(external=[fake_snap("one", disks=("vda",))])
[entry] = p.get_snapshot(EXTERNAL)
self.assertEqual(entry["detaches"], [{"dev": "vdb", "path": "/pool/later.qcow2"}])
def test_detaches_comes_from_the_persistent_definition(self):
# vdc was hot-plugged into the running VM only; vdb waits for the next start
live = DOMAIN.replace("/pool/later.qcow2", "/pool/live-only.qcow2").replace("dev='vdb'", "dev='vdc'")
p = proxy(external=[fake_snap("one")])
p._XMLDesc = lambda flags: DOMAIN if flags == PERSISTENT_XML else live
[entry] = p.get_snapshot(EXTERNAL)
self.assertEqual(entry["detaches"], [{"dev": "vdb", "path": "/pool/later.qcow2"}])
def test_lists_are_by_kind(self):
p = proxy(internal=[fake_snap("i")], external=[fake_snap("e")])
self.assertEqual([s["name"] for s in p.get_snapshot(INTERNAL)], ["i"])
self.assertEqual([s["name"] for s in p.get_snapshot(EXTERNAL)], ["e"])
class SnapshotJobsTestCase(unittest.TestCase):
def test_running_jobs_are_reported_per_disk(self):
p = proxy(jobs={"vda": {"cur": 5, "end": 10, "type": 3, "bandwidth": 0}})
self.assertEqual(p.get_snapshot_jobs(), [{"dev": "vda", "cur": 5, "end": 10}])
def test_a_stopped_vm_has_no_jobs(self):
p = proxy(status=5)
self.assertEqual(p.get_snapshot_jobs(), [])
p.instance.blockJobInfo.assert_not_called()
class SnapshotLookupTestCase(unittest.TestCase):
def test_a_name_of_the_other_kind_is_refused(self):
p = proxy(internal=[fake_snap("i")], external=[fake_snap("e")])
with self.assertRaises(util.OperationError):
p._snapshot("i", external=True)
with self.assertRaises(util.OperationError):
p._snapshot("e", external=False)
self.assertEqual(p._snapshot("e", external=True).getName(), "e")
class SnapshotStateTestCase(unittest.TestCase):
def test_managed_save_is_refused(self):
with self.assertRaisesRegex(util.OperationError, "saved state"):
proxy(managed_save=True)._check_snapshot_state()
def test_running_job_is_refused(self):
with self.assertRaisesRegex(util.OperationError, "still running"):
proxy(jobs={"vda": {"cur": 1, "end": 2}})._check_snapshot_state()
def test_a_mixed_vm_passes_so_it_can_be_cleaned_up(self):
proxy(internal=[fake_snap("i")], external=[fake_snap("e")])._check_snapshot_state()
def test_a_clean_vm_passes(self):
proxy()._check_snapshot_state()
DISK_ONLY_ATOMIC = libvirt.VIR_DOMAIN_SNAPSHOT_CREATE_DISK_ONLY | libvirt.VIR_DOMAIN_SNAPSHOT_CREATE_ATOMIC
def created(p):
xml, flags = p._snapshotCreateXML.call_args[0]
return xml, flags
class CreateTestCase(unittest.TestCase):
def test_disk_only_snapshot(self):
p = proxy()
p.create_external_snapshot("snap-1", "desc")
xml, flags = created(p)
self.assertEqual(flags, DISK_ONLY_ATOMIC)
self.assertIn("<name>snap-1</name>", xml)
self.assertNotIn("<memory", xml)
p.refresh_instance_pools.assert_called_once()
def test_description_is_escaped(self):
p = proxy()
p.create_external_snapshot("s", "<x> & 'y'")
xml, _ = created(p)
self.assertIn("&lt;x&gt; &amp;", xml)
self.assertNotIn("<x>", xml)
def test_memory_snapshot_goes_into_the_first_disks_pool(self):
p = proxy()
p.get_volume_by_path = MagicMock()
pool = p.get_volume_by_path.return_value.storagePoolLookupByVolume.return_value
pool.info.return_value = [2, 0, 0, 2 * 1024**3]
p.create_external_snapshot("m", memory=True)
xml, flags = created(p)
p.get_volume_by_path.assert_called_once_with("/pool/a.qcow2")
self.assertEqual(flags, libvirt.VIR_DOMAIN_SNAPSHOT_CREATE_ATOMIC)
self.assertIn(
"<memory snapshot='external' file='/pool/11111111-2222-3333-4444-555555555555.m.mem'/>", xml
)
def refused(self, p, pattern, **kwargs):
with self.assertRaisesRegex(util.OperationError, pattern):
p.create_external_snapshot(kwargs.pop("name", "s"), **kwargs)
p._snapshotCreateXML.assert_not_called()
def test_invalid_names_are_refused(self):
for name in ("", ".hidden", "a/b", "a b", "x" * 65, "<x>"):
with self.subTest(name=name):
self.refused(proxy(), "Snapshot names", name=name)
def test_names_are_checked_only_on_create(self):
# snapshots made by the old code are named s1.<name> and stay usable
p = proxy(external=[fake_snap("s1.old", current=True)])
p.instance.revertToSnapshot = MagicMock()
p.revert_external_snapshot("s1.old")
p.instance.revertToSnapshot.assert_called_once()
def test_a_taken_name_is_refused(self):
self.refused(proxy(external=[fake_snap("s", current=True)]), "already exists")
def test_shared_state_checks_apply(self):
self.refused(proxy(managed_save=True), "saved state")
def test_internal_snapshots_are_refused(self):
self.refused(proxy(internal=[fake_snap("i")]), "internal snapshots")
def test_newer_snapshots_after_a_revert_are_refused(self):
self.refused(proxy(external=[fake_snap("old", current=True, children=1)]), "Newer snapshots")
def test_snapshots_without_a_current_one_are_refused(self):
# a create would start a second root
self.refused(proxy(external=[fake_snap("old")]), "revert to a snapshot first")
def test_a_disk_that_is_not_a_file_is_refused(self):
xml = "<domain><devices><disk type='block' device='disk'><source dev='/dev/vg/x'/><target dev='vda'/></disk></devices></domain>"
self.refused(proxy(xml=xml), "vda is not a file")
def test_cdrom_with_media_is_refused(self):
xml = DOMAIN.replace("<target dev='sda'", "<source file='/iso/x.iso'/><target dev='sda'")
self.refused(proxy(xml=xml), "Eject")
def test_memory_needs_a_running_vm(self):
self.refused(proxy(status=3), "running VM", memory=True)
self.refused(proxy(status=5), "running VM", memory=True)
def test_memory_needs_a_pool(self):
p = proxy()
p.get_volume_by_path = MagicMock(side_effect=libvirtError("no volume"))
self.refused(p, "not in a storage pool", memory=True)
def test_memory_needs_room_in_the_pool(self):
p = proxy()
p.get_volume_by_path = MagicMock()
pool = p.get_volume_by_path.return_value.storagePoolLookupByVolume.return_value
pool.info.return_value = [2, 0, 0, 512 * 1024**2]
self.refused(p, "less free space", memory=True)
REVERT_RUNNING = libvirt.VIR_DOMAIN_SNAPSHOT_REVERT_RUNNING
REVERT_PAUSED = libvirt.VIR_DOMAIN_SNAPSHOT_REVERT_PAUSED
class RevertTestCase(unittest.TestCase):
def reverted_with(self, status, memory=False):
snap = fake_snap("s", memory=memory, current=True)
p = proxy(status=status, external=[snap])
p.revert_external_snapshot("s")
p.instance.revertToSnapshot.assert_called_once_with(snap, unittest.mock.ANY)
p.refresh_instance_pools.assert_called_once()
return p.instance.revertToSnapshot.call_args[0][1]
def test_disk_only_flags_keep_the_vm_state(self):
self.assertEqual(self.reverted_with(1), REVERT_RUNNING)
self.assertEqual(self.reverted_with(3), REVERT_PAUSED)
self.assertEqual(self.reverted_with(5), 0)
def test_memory_snapshot_resumes_as_saved(self):
self.assertEqual(self.reverted_with(1, memory=True), 0)
self.assertEqual(self.reverted_with(5, memory=True), 0)
def test_shared_state_checks_apply(self):
p = proxy(external=[fake_snap("s")], managed_save=True)
with self.assertRaisesRegex(util.OperationError, "saved state"):
p.revert_external_snapshot("s")
p.instance.revertToSnapshot.assert_not_called()
def collision(self):
err = libvirtError("external snapshot file for disk vda already exists and is not a block device")
err.get_error_code = lambda: libvirt.VIR_ERR_CONFIG_UNSUPPORTED
return err
def test_same_second_revert_collision_is_retried_once(self):
# libvirt has already stopped a running VM when the overlay names
# collide; the retry a second later finishes the revert.
snap = fake_snap("s")
p = proxy(external=[snap])
p.instance.revertToSnapshot.side_effect = [self.collision(), None]
with unittest.mock.patch("vrtManager.instance.time.sleep") as sleep:
p.revert_external_snapshot("s")
sleep.assert_called_once_with(1)
self.assertEqual(
p.instance.revertToSnapshot.call_args_list,
[unittest.mock.call(snap, REVERT_RUNNING)] * 2,
)
def test_a_second_collision_is_reported_with_the_vm_state(self):
p = proxy(external=[fake_snap("s")])
p.instance.revertToSnapshot.side_effect = [self.collision(), self.collision()]
p.instance.isActive.return_value = False
with unittest.mock.patch("vrtManager.instance.time.sleep"):
with self.assertRaisesRegex(util.OperationError, "try again in a moment.*shut off"):
p.revert_external_snapshot("s")
def test_other_libvirt_errors_pass_through(self):
p = proxy(external=[fake_snap("s")])
p.instance.revertToSnapshot.side_effect = libvirtError("boom")
with self.assertRaises(libvirtError) as ctx:
p.revert_external_snapshot("s")
self.assertNotIsInstance(ctx.exception, util.OperationError)
def test_an_old_uefi_snapshot_saved_with_a_rom_loader_is_refused(self):
# The old code switched pflash to rom while taking the snapshot, so
# the saved definition has rom; reverting would leave the VM unbootable.
uefi = DOMAIN.replace("<devices>", "<os><loader readonly='yes' type='pflash'>/ovmf/CODE.fd</loader></os><devices>")
snap = fake_snap("s1.old", current=True)
snap.getXMLDesc.return_value = snapshot_xml("s1.old").replace(
"<domain>", "<domain><os><loader readonly='yes' type='rom'>/ovmf/CODE.fd</loader></os>"
)
p = proxy(external=[snap], xml=uefi)
with self.assertRaisesRegex(util.OperationError, "older version"):
p.revert_external_snapshot("s1.old")
p.instance.revertToSnapshot.assert_not_called()
def test_an_old_running_uefi_snapshot_is_judged_by_its_inactive_definition(self):
# a running VM's snapshot keeps the live definition (pflash) and the
# persistent one (rom); the revert restores the persistent one
uefi = DOMAIN.replace("<devices>", "<os><loader readonly='yes' type='pflash'>/ovmf/CODE.fd</loader></os><devices>")
snap = fake_snap("s1.old", current=True)
snap.getXMLDesc.return_value = snapshot_xml("s1.old").replace(
"<domain>", "<domain><os><loader readonly='yes' type='pflash'>/ovmf/CODE.fd</loader></os>"
).replace(
"</domainsnapshot>",
"<inactiveDomain><os><loader readonly='yes' type='rom'>/ovmf/CODE.fd</loader></os>"
"<devices/></inactiveDomain></domainsnapshot>",
)
p = proxy(external=[snap], xml=uefi)
with self.assertRaisesRegex(util.OperationError, "older version"):
p.revert_external_snapshot("s1.old")
p.instance.revertToSnapshot.assert_not_called()
def test_a_failed_retry_still_reports_the_vm_state(self):
# the collision already stopped the VM; another error on the retry
# must not hide that
p = proxy(external=[fake_snap("s")])
p.instance.revertToSnapshot.side_effect = [self.collision(), libvirtError("disk busy")]
p.instance.isActive.return_value = False
with unittest.mock.patch("vrtManager.instance.time.sleep"):
with self.assertRaisesRegex(util.OperationError, "disk busy.*The VM is shut off"):
p.revert_external_snapshot("s")
class DeleteTestCase(unittest.TestCase):
def test_delete_is_one_libvirt_call(self):
snap = fake_snap("s", current=True)
p = proxy(external=[snap])
p.delete_external_snapshot("s")
snap.delete.assert_called_once_with(0)
p.refresh_instance_pools.assert_called_once()
def test_running_job_is_refused(self):
snap = fake_snap("s")
p = proxy(external=[snap], jobs={"vda": {"cur": 1, "end": 2}})
with self.assertRaisesRegex(util.OperationError, "still running"):
p.delete_external_snapshot("s")
snap.delete.assert_not_called()
CHAIN = """<domain><devices>
<disk type='file' device='disk'><source file='/pool/a.two'/><target dev='vda'/>
<backingStore type='file'><source file='/pool/a.one'/>
<backingStore type='file'><source file='/pool/a.qcow2'/><backingStore/></backingStore>
</backingStore>
</disk>
</devices></domain>"""
def chain_snap(name, overlay, memory=None):
snap = fake_snap(name)
mem = f"<memory snapshot='external' file='{memory}'/>" if memory else "<memory snapshot='no'/>"
snap.getXMLDesc.return_value = (
f"<domainsnapshot><name>{name}</name><creationTime>1700000000</creationTime>{mem}"
f"<disks><disk name='vda' snapshot='external'><source file='{overlay}'/></disk></disks>"
"<domain><devices/></domain></domainsnapshot>"
)
return snap
class LeftBehindTestCase(unittest.TestCase):
"""libvirt only logs a warning when it cannot remove a merged or
discarded file (on NFS it removes them as the qemu user), so the delete
and revert report the files nothing uses any more that still exist."""
def run_delete(self, still_there):
one = chain_snap("one", "/pool/a.one", memory="/pool/vm.one.mem")
two = chain_snap("two", "/pool/a.two")
p = proxy(external=[one, two])
after = CHAIN.replace("<source file='/pool/a.one'/>", "<source file='/pool/a.qcow2'/>").replace(
"<backingStore type='file'><source file='/pool/a.qcow2'/><backingStore/></backingStore>", "<backingStore/>"
)
p.instance.XMLDesc.side_effect = [CHAIN, after]
p.instance.listAllSnapshots.side_effect = [[one, two], [two]]
def volume(path):
if path not in still_there:
raise libvirtError("no volume")
return MagicMock()
p.get_volume_by_path = volume
return p.delete_external_snapshot("one")
def test_files_libvirt_could_not_remove_are_reported(self):
self.assertEqual(
self.run_delete({"/pool/a.one", "/pool/vm.one.mem", "/pool/a.two", "/pool/a.qcow2"}),
["/pool/a.one", "/pool/vm.one.mem"],
)
def test_nothing_is_reported_when_libvirt_removed_them(self):
# files still in use (base, the newer overlay) are never reported
self.assertEqual(self.run_delete({"/pool/a.two", "/pool/a.qcow2"}), [])
def revert(self, before, after, backing):
"""Revert to one; backing maps each file still on disk to its backing file."""
one = chain_snap("one", "/pool/a.one")
p = proxy(external=[one])
p.instance.XMLDesc.side_effect = [before, after]
p.instance.listAllSnapshots.side_effect = [[one], [one]]
def volume(path):
if path not in backing:
raise libvirtError("no volume")
parent = f"<backingStore><path>{backing[path]}</path></backingStore>" if backing[path] else ""
return MagicMock(**{"XMLDesc.return_value": f"<volume><target><path>{path}</path></target>{parent}</volume>"})
p.get_volume_by_path = volume
return p.revert_external_snapshot("one")
def test_revert_reports_a_discarded_overlay_left_on_disk(self):
before = CHAIN.replace("/pool/a.two", "/pool/a.after")
after = CHAIN.replace("/pool/a.two", "/pool/a.1700000001")
left = self.revert(before, after, {"/pool/a.after": "/pool/a.one", "/pool/a.1700000001": "/pool/a.one"})
self.assertEqual(left, ["/pool/a.after"])
def test_a_shut_off_vm_overlay_is_found_through_the_storage(self):
# a shut-off VM's definition does not show the backing chain
def plain(top):
return f"<domain><devices><disk type='file' device='disk'><source file='{top}'/><target dev='vda'/></disk></devices></domain>"
left = self.revert(plain("/pool/a.after"), plain("/pool/a.1700000001"), {"/pool/a.after": "/pool/a.one"})
self.assertEqual(left, ["/pool/a.after"])
def test_revert_never_reports_a_detached_disk(self):
# vdb and vdc were added after the snapshot: the revert detaches them
# and keeps their files on purpose, also when vdc is itself an overlay
later = (
"<disk type='file' device='disk'><source file='/pool/later.qcow2'/><target dev='vdb'/></disk>"
"<disk type='file' device='disk'><source file='/pool/clone.qcow2'/><target dev='vdc'/>"
"<backingStore type='file'><source file='/templates/base.qcow2'/><backingStore/></backingStore></disk>"
)
before = CHAIN.replace("</devices>", later + "</devices>")
left = self.revert(before, CHAIN, {"/pool/later.qcow2": None, "/pool/clone.qcow2": "/templates/base.qcow2"})
self.assertEqual(left, [])
def test_revert_never_reports_a_detached_overlay_of_a_snapshot_file(self):
linked = (
"<disk type='file' device='disk'><source file='/pool/linked.qcow2'/><target dev='vdb'/>"
"<backingStore type='file'><source file='/pool/a.one'/><backingStore/></backingStore></disk>"
)
before = CHAIN.replace("</devices>", linked + "</devices>")
left = self.revert(before, CHAIN, {"/pool/linked.qcow2": "/pool/a.one"})
self.assertEqual(left, [])
class KindTestCase(unittest.TestCase):
def test_internal_endpoints_refuse_external_names(self):
snap = fake_snap("e")
p = proxy(external=[snap])
for call in (p.snapshot_revert, p.snapshot_delete):
with self.subTest(call.__name__), self.assertRaises(util.OperationError):
call("e")
snap.delete.assert_not_called()
p.instance.revertToSnapshot.assert_not_called()
def test_external_endpoints_refuse_internal_names(self):
snap = fake_snap("i")
p = proxy(internal=[snap])
for call in (p.revert_external_snapshot, p.delete_external_snapshot):
with self.subTest(call.__name__), self.assertRaises(util.OperationError):
call("i")
snap.delete.assert_not_called()
def test_internal_create_refuses_a_vm_with_external_snapshots(self):
p = proxy(external=[fake_snap("e")])
with self.assertRaisesRegex(util.OperationError, "external snapshots"):
p.create_snapshot("i")
p._snapshotCreateXML.assert_not_called()
if __name__ == "__main__":
unittest.main()