OpenStack Compute (Nova)
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

943 lines
36 KiB

# Copyright 2010 United States Government as represented by the
# Administrator of the National Aeronautics and Space Administration.
# All Rights Reserved.
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
"""Base classes for our unit tests.
Allows overriding of flags for use of fakes, and some black magic for
inline callbacks.
import nova.monkey_patch # noqa
import abc
import builtins
import collections
import contextlib
import copy
import datetime
import inspect
import itertools
import os
import os.path
import pprint
import sys
import fixtures
import mock
from oslo_cache import core as cache
from oslo_concurrency import lockutils
from oslo_config import cfg
from oslo_config import fixture as config_fixture
from oslo_log.fixture import logging_error as log_fixture
from oslo_log import log as logging
from oslo_serialization import jsonutils
from oslo_utils.fixture import uuidsentinel as uuids
from oslo_utils import timeutils
from oslo_versionedobjects import fixture as ovo_fixture
from oslotest import base
from oslotest import mock_fixture
from sqlalchemy.dialects import sqlite
import testtools
from nova.api.openstack import wsgi_app
from nova.compute import rpcapi as compute_rpcapi
from nova import context
from nova.db.sqlalchemy import api as sqlalchemy_api
from nova import exception
from nova import objects
from nova.objects import base as objects_base
from nova import quota
from nova.tests import fixtures as nova_fixtures
from nova.tests.unit import conf_fixture
from nova.tests.unit import matchers
from nova.tests.unit import policy_fixture
from nova import utils
from nova.virt import images
CONF.set_override('use_stderr', False)
logging.setup(CONF, 'nova')
LOG = logging.getLogger(__name__)
_TRUE_VALUES = ('True', 'true', '1', 'yes')
CELL1_NAME = 'cell1'
# For compatibility with the large number of tests which use test.nested
nested = utils.nested_contexts
class TestingException(Exception):
# NOTE(claudiub): this needs to be called before any mock.patch calls are
# being done, and especially before any other test classes load. This fixes
# the mock.patch autospec issue:
def _poison_unfair_compute_resource_semaphore_locking():
"""Ensure that every locking on COMPUTE_RESOURCE_SEMAPHORE is called with
orig_synchronized = utils.synchronized
def poisoned_synchronized(*args, **kwargs):
# Only check fairness if the decorator is used with
# COMPUTE_RESOURCE_SEMAPHORE. But the name of the semaphore can be
# passed as args or as kwargs.
# Note that we cannot import COMPUTE_RESOURCE_SEMAPHORE as that would
# apply the decorators we want to poison here.
if len(args) >= 1:
name = args[0]
name = kwargs.get("name")
if name == "compute_resources" and not kwargs.get("fair", False):
raise AssertionError(
'Locking on COMPUTE_RESOURCE_SEMAPHORE should always be fair. '
'See bug 1864122.')
# go and act like the original decorator
return orig_synchronized(*args, **kwargs)
# replace the synchronized decorator factory with our own that checks the
# params passed in
utils.synchronized = poisoned_synchronized
# NOTE(gibi): This poisoning needs to be done in import time as decorators are
# applied in import time on the ResourceTracker
class NovaExceptionReraiseFormatError(object):
real_log_exception = exception.NovaException._log_exception
def patch(cls):
exception.NovaException._log_exception = cls._wrap_log_exception
def _wrap_log_exception(self):
exc_info = sys.exc_info()
raise exc_info[1]
# NOTE(melwitt) This needs to be done at import time in order to also catch
# NovaException format errors that are in mock decorators. In these cases, the
# errors will be raised during test listing, before tests actually run.
class TestCase(base.BaseTestCase):
"""Test case base class for all unit tests.
Due to the slowness of DB access, please consider deriving from
`NoDBTestCase` first.
# USES_DB is set to False for tests that inherit from NoDBTestCase.
USES_DB = True
# USES_DB_SELF is set to True in tests that specifically want to use the
# database but need to configure it themselves, for example to setup the
# API DB but not the cell DB. In those cases the test will override
# USES_DB_SELF = True but inherit from the NoDBTestCase class so it does
# not get the default fixture setup when using a database (which is the
# API and cell DBs, and adding the default flavors).
# Setting to True makes the test use the RPCFixture.
# The number of non-cell0 cells to create. This is only used in the
# base class when USES_DB is True.
def setUp(self):
"""Run before each test method to initialize test environment."""
# Ensure BaseTestCase's ConfigureLogging fixture is disabled since
# we're using our own (StandardLogging).
with fixtures.EnvironmentVariable('OS_LOG_CAPTURE', '0'):
super(TestCase, self).setUp()
# How many of which service we've started. {$service-name: $count}
self._service_fixture_count = collections.defaultdict(int)
self.stdlog = self.useFixture(nova_fixtures.StandardLogging())
# NOTE(sdague): because of the way we were using the lock
# wrapper we ended up with a lot of tests that started
# relying on global external locking being set up for them. We
# consider all of these to be *bugs*. Tests should not require
# global external locking, or if they do, they should
# explicitly set it up themselves.
# The following REQUIRES_LOCKING class parameter is provided
# as a bridge to get us there. No new tests should be added
# that require it, and existing classes and tests should be
# fixed to not need it.
lock_path = self.useFixture(fixtures.TempDir()).path
self.fixture = self.useFixture(
if self.STUB_RPC:
# we cannot set this in the ConfFixture as oslo only registers the
# notification opts at the first instantiation of a Notifier that
# happens only in the RPCFixture
CONF.set_default('driver', ['test'],
# NOTE(danms): Make sure to reset us back to non-remote objects
# for each test to avoid interactions. Also, backup the object
# registry.
objects_base.NovaObject.indirection_api = None
self._base_test_obj_backup = copy.copy(
# NOTE(danms): Reset the cached list of cells
from nova.compute import api
api.CELLS = []
context.CELL_CACHE = {}
context.CELLS = []
self.computes = {}
self.cell_mappings = {}
self.host_mappings = {}
# NOTE(danms): If the test claims to want to set up the database
# itself, then it is responsible for all the mapping stuff too.
if self.USES_DB:
# NOTE(danms): Full database setup involves a cell0, cell1,
# and the relevant mappings.
elif not self.USES_DB_SELF:
# NOTE(danms): If not using the database, we mock out the
# mapping stuff and effectively collapse everything to a
# single cell.
# NOTE(blk-u): WarningsFixture must be after the Database fixture
# because sqlalchemy-migrate messes with the warnings filters.
# Reset the global QEMU version flag.
images.QEMU_VERSION = None
# Reset the compute RPC API globals (mostly the _ROUTER).
self.policy = self.useFixture(policy_fixture.PolicyFixture())
# NOTE(mikal): make sure we don't load a privsep helper accidentally
# FIXME(danms): Disable this for all tests by default to avoid breaking
# any that depend on default/previous ordering
# NOTE(melwitt): Reset the cached set of projects
# make sure that the wsgi app is fully initialized for all testcase
# instead of only once initialized for test worker
def _setup_cells(self):
"""Setup a normal cellsv2 environment.
This sets up the CellDatabase fixture with two cells, one cell0
and one normal cell. CellMappings are created for both so that
cells-aware code can find those two databases.
celldbs = nova_fixtures.CellDatabases()
ctxt = context.get_context()
fake_transport = 'fake://nowhere/'
c0 = objects.CellMapping(
self.cell_mappings[] = c0
for x in range(self.NUMBER_OF_CELLS):
name = 'cell%i' % (x + 1)
uuid = getattr(uuids, name)
cell = objects.CellMapping(
self.cell_mappings[name] = cell
# cell1 is the default cell
celldbs.add_cell_database(uuid, default=(x == 0))
def _restore_obj_registry(self):
objects_base.NovaObjectRegistry._registry._obj_classes = \
def _clear_attrs(self):
# Delete attributes that don't start with _ so they don't pin
# memory around unnecessarily for the duration of the test
# suite
for key in [k for k in self.__dict__.keys() if k[0] != '_']:
# NOTE(gmann): Skip attribute 'id' because if tests are being
# generated using testscenarios then, 'id' attribute is being
# added during cloning the tests. And later that 'id' attribute
# is being used by test suite to generate the results for each
# newly generated tests by testscenarios.
if key != 'id':
del self.__dict__[key]
def stub_out(self, old, new):
"""Replace a function for the duration of the test.
Use the monkey patch fixture to replace a function for the
duration of a test. Useful when you want to provide fake
methods instead of mocks during testing.
self.useFixture(fixtures.MonkeyPatch(old, new))
def patch_exists(patched_path, result):
"""Provide a static method version of patch_exists(), which if you
haven't already imported nova.test can be slightly easier to
use as a context manager within a test method via:
def test_something(self):
with self.patch_exists(path, True):
return patch_exists(patched_path, result)
def patch_open(patched_path, read_data):
"""Provide a static method version of patch_open() which is easier to
use as a context manager within a test method via:
def test_something(self):
with self.patch_open(path, "fake contents of file"):
return patch_open(patched_path, read_data)
def flags(self, **kw):
"""Override flag variables for a test."""
group = kw.pop('group', None)
for k, v in kw.items():
CONF.set_override(k, v, group)
def enforce_fk_constraints(self, engine=None):
if engine is None:
engine = sqlalchemy_api.get_engine()
dialect = engine.url.get_dialect()
if dialect == sqlite.dialect:
# We're seeing issues with foreign key support in SQLite 3.6.20
# SQLAlchemy doesn't support it at all with < SQLite 3.6.19
# It works fine in SQLite 3.7.
# So return early to skip this test if running SQLite < 3.7
import sqlite3
tup = sqlite3.sqlite_version_info
if tup[0] < 3 or (tup[0] == 3 and tup[1] < 7):
'sqlite version too old for reliable SQLA foreign_keys')
engine.connect().execute("PRAGMA foreign_keys = ON")
def start_service(self, name, host=None, cell_name=None, **kwargs):
# Disallow starting multiple scheduler services
if name == 'scheduler' and self._service_fixture_count[name]:
raise TestingException("Duplicate start_service(%s)!" % name)
cell = None
# if the host is None then the remains defaulted to
# 'fake-mini' (originally done in ConfFixture)
if host is not None:
# Make sure that is relevant to the right hostname
if name == 'compute' and self.USES_DB:
# NOTE(danms): We need to create the HostMapping first, because
# otherwise we'll fail to update the scheduler while running
# the compute node startup routines below.
ctxt = context.get_context()
cell_name = cell_name or CELL1_NAME
cell = self.cell_mappings[cell_name]
if (host or name) not in self.host_mappings:
# NOTE(gibi): If the HostMapping does not exists then this is
# the first start of the service so we create the mapping.
hm = objects.HostMapping(context=ctxt,
host=host or name,
self.host_mappings[] = hm
svc = self.useFixture(
nova_fixtures.ServiceFixture(name, host, cell=cell, **kwargs))
# Keep track of how many instances of this service are running.
self._service_fixture_count[name] += 1
real_stop = svc.service.stop
# Make sure stopping the service decrements the active count, so that
# start,stop,start doesn't trigger the "Duplicate start_service"
# exception.
def patch_stop(*a, **k):
self._service_fixture_count[name] -= 1
return real_stop(*a, **k)
svc.service, 'stop', patch_stop))
return svc.service
def _start_compute(self, host, cell_name=None):
"""Start a nova compute service on the given host
:param host: the name of the host that will be associated to the
compute service.
:param cell_name: optional name of the cell in which to start the
compute service
:return: the nova compute service object
compute = self.start_service('compute', host=host, cell_name=cell_name)
self.computes[host] = compute
return compute
def _run_periodics(self):
"""Run the update_available_resource task on every compute manager
This runs periodics on the computes in an undefined order; some child
class redefine this function to force a specific order.
ctx = context.get_admin_context()
for host, compute in self.computes.items():'Running periodic for compute (%s)', host)
# Make sure the context is targeted to the proper cell database
# for multi-cell tests.
with context.target_cell(
ctx, self.host_mappings[host].cell_mapping) as cctxt:
compute.manager.update_available_resource(cctxt)'Finished with periodics')
def restart_compute_service(self, compute, keep_hypervisor_state=True):
"""Stops the service and starts a new one to have realistic restart
:param:compute: the nova-compute service to be restarted
:param:keep_hypervisor_state: If true then already defined instances
will survive the compute service restart.
If false then the new service will see
an empty hypervisor
:returns: a new compute service instance serving the same host and
and node
# NOTE(gibi): The service interface cannot be used to simulate a real
# service restart as the manager object will not be recreated after a
# service.stop() and service.start() therefore the manager state will
# survive. For example the resource tracker will not be recreated after
# a stop start. The service.kill() call cannot help as it deletes
# the service from the DB which is unrealistic and causes that some
# operation that refers to the killed host (e.g. evacuate) fails.
# So this helper method will stop the original service and then starts
# a brand new compute service for the same host and node. This way
# a new ComputeManager instance will be created and initialized during
# the service startup.
# this service was running previously so we have to make sure that
# we restart it in the same cell
cell_name = self.host_mappings[]
if keep_hypervisor_state:
# NOTE(gibi): FakeDriver does not provide a meaningful way to
# define some servers that exists already on the hypervisor when
# the driver is (re)created during the service startup. This means
# that we cannot simulate that the definition of a server
# survives a nova-compute service restart on the hypervisor.
# Instead here we save the FakeDriver instance that knows about
# the defined servers and inject that driver into the new Manager
# class during the startup of the compute service.
old_driver = compute.manager.driver
with mock.patch(
'nova.virt.driver.load_compute_driver') as load_driver:
load_driver.return_value = old_driver
new_compute = self.start_service(
'compute',, cell_name=cell_name)
new_compute = self.start_service(
'compute',, cell_name=cell_name)
return new_compute
def assertJsonEqual(self, expected, observed, message=''):
"""Asserts that 2 complex data structures are json equivalent.
We use data structures which serialize down to json throughout
the code, and often times we just need to know that these are
json equivalent. This means that list order is not important,
and should be sorted.
Because this is a recursive set of assertions, when failure
happens we want to expose both the local failure and the
global view of the 2 data structures being compared. So a
MismatchError which includes the inner failure as the
mismatch, and the passed in expected / observed as matchee /
if isinstance(expected, str):
expected = jsonutils.loads(expected)
if isinstance(observed, str):
observed = jsonutils.loads(observed)
def sort_key(x):
if isinstance(x, (set, list)) or isinstance(x, datetime.datetime):
return str(x)
if isinstance(x, dict):
items = ((sort_key(key), sort_key(value))
for key, value in x.items())
return sorted(items)
return x
def inner(expected, observed, path='root'):
if isinstance(expected, dict) and isinstance(observed, dict):
len(expected), len(observed),
('path: %s. Different dict key sets\n'
'difference=%s') %
expected_keys = sorted(expected)
observed_keys = sorted(observed)
expected_keys, observed_keys,
'path: %s. Dict keys are not equal' % path)
for key in expected:
inner(expected[key], observed[key], path + '.%s' % key)
elif (isinstance(expected, (list, tuple, set)) and
isinstance(observed, (list, tuple, set))):
len(expected), len(observed),
('path: %s. Different list items\n'
'difference=%s') %
sorted(expected, key=sort_key),
sorted(observed, key=sort_key),
[a for a in itertools.chain(expected, observed) if
(a not in expected) or (a not in observed)]))
expected_values_iter = iter(sorted(expected, key=sort_key))
observed_values_iter = iter(sorted(observed, key=sort_key))
for i in range(len(expected)):
next(observed_values_iter), path + '[%s]' % i)
self.assertEqual(expected, observed, 'path: %s' % path)
inner(expected, observed)
except testtools.matchers.MismatchError as e:
difference = e.mismatch.describe()
if message:
message = 'message: %s\n' % message
msg = "\nexpected:\n%s\nactual:\n%s\ndifference:\n%s\n%s" % (
error = AssertionError(msg)
error.expected = expected
error.observed = observed
error.difference = difference
raise error
def assertXmlEqual(self, expected, observed, **options):
self.assertThat(observed, matchers.XMLMatches(expected, **options))
def assertPublicAPISignatures(self, baseinst, inst):
def get_public_apis(inst):
methods = {}
def findmethods(object):
return inspect.ismethod(object) or inspect.isfunction(object)
for (name, value) in inspect.getmembers(inst, findmethods):
if name.startswith("_"):
methods[name] = value
return methods
baseclass = baseinst.__class__.__name__
basemethods = get_public_apis(baseinst)
implmethods = get_public_apis(inst)
extranames = []
for name in sorted(implmethods.keys()):
if name not in basemethods:
self.assertEqual([], extranames,
"public APIs not listed in base class %s" %
for name in sorted(implmethods.keys()):
# NOTE(stephenfin): We ignore type annotations
baseargs = utils.getargspec(basemethods[name])[:-1]
implargs = utils.getargspec(implmethods[name])[:-1]
self.assertEqual(baseargs, implargs,
"%s args don't match base class %s" %
(name, baseclass))
class APICoverage(object):
cover_api = None
def test_api_methods(self):
api_methods = [x for x in dir(self.cover_api)
if not x.startswith('_')]
test_methods = [x[5:] for x in dir(self)
if x.startswith('test_')]
class SubclassSignatureTestCase(testtools.TestCase, metaclass=abc.ABCMeta):
"""Ensure all overridden methods of all subclasses of the class
under test exactly match the signature of the base class.
A subclass of SubclassSignatureTestCase should define a method
_get_base_class which:
* Returns a base class whose subclasses will all be checked
* Ensures that all subclasses to be tested have been imported
SubclassSignatureTestCase defines a single test, test_signatures,
which does a recursive, depth-first check of all subclasses, ensuring
that their method signatures are identical to those of the base class.
def _get_base_class(self):
raise NotImplementedError()
def setUp(self):
self.base = self._get_base_class()
super(SubclassSignatureTestCase, self).setUp()
def _get_argspecs(cls):
"""Return a dict of method_name->argspec for every method of cls."""
argspecs = {}
# getmembers returns all members, including members inherited from
# the base class. It's redundant for us to test these, but as
# they'll always pass it's not worth the complexity to filter them out.
for (name, method) in inspect.getmembers(cls, inspect.ismethod):
# Subclass __init__ methods can usually be legitimately different
if name == '__init__':
while hasattr(method, '__wrapped__'):
# This is a wrapped function. The signature we're going to
# see here is that of the wrapper, which is almost certainly
# going to involve varargs and kwargs, and therefore is
# unlikely to be what we want. If the wrapper manupulates the
# arguments taken by the wrapped function, the wrapped function
# isn't what we want either. In that case we're just stumped:
# if it ever comes up, add more knobs here to work round it (or
# stop using a dynamic language).
# Here we assume the wrapper doesn't manipulate the arguments
# to the wrapped function and inspect the wrapped function
# instead.
method = getattr(method, '__wrapped__')
argspecs[name] = utils.getargspec(method)
return argspecs
def _clsname(cls):
"""Return the fully qualified name of cls."""
return "%s.%s" % (cls.__module__, cls.__name__)
def _test_signatures_recurse(self, base, base_argspecs):
for sub in base.__subclasses__():
sub_argspecs = self._get_argspecs(sub)
# Check that each subclass method matches the signature of the
# base class
for (method, sub_argspec) in sub_argspecs.items():
# Methods which don't override methods in the base class
# are good.
if method in base_argspecs:
self.assertEqual(base_argspecs[method], sub_argspec,
'Signature of %(sub)s.%(method)s '
'differs from superclass %(base)s' %
{'base': self._clsname(base),
'sub': self._clsname(sub),
'method': method})
# Recursively check this subclass
self._test_signatures_recurse(sub, sub_argspecs)
def test_signatures(self):
self._test_signatures_recurse(self.base, self._get_argspecs(self.base))
class TimeOverride(fixtures.Fixture):
"""Fixture to start and remove time override."""
def setUp(self):
super(TimeOverride, self).setUp()
class NoDBTestCase(TestCase):
"""`NoDBTestCase` differs from TestCase in that DB access is not supported.
This makes tests run significantly faster. If possible, all new tests
should derive from this class.
USES_DB = False
class MatchType(object):
"""Matches any instance of a specified type
The MatchType class is a helper for use with the mock.assert_called_with()
method that lets you assert that a particular parameter has a specific data
type. It enables stricter checking than the built in mock.ANY helper.
Example usage could be:
def __init__(self, wanttype):
self.wanttype = wanttype
def __eq__(self, other):
return type(other) == self.wanttype
def __ne__(self, other):
return type(other) != self.wanttype
def __repr__(self):
return "<MatchType:" + str(self.wanttype) + ">"
class MatchObjPrims(object):
"""Matches objects with equal primitives."""
def __init__(self, want_obj):
self.want_obj = want_obj
def __eq__(self, other):
return objects_base.obj_equal_prims(other, self.want_obj)
def __ne__(self, other):
return not other == self.want_obj
def __repr__(self):
return '<MatchObjPrims:' + str(self.want_obj) + '>'
class ContainKeyValue(object):
"""Checks whether a key/value pair is in a dict parameter.
The ContainKeyValue class is a helper for use with the mock.assert_*()
method that lets you assert that a particular dict contain a key/value
pair. It enables stricter checking than the built in mock.ANY helper.
Example usage could be:
ContainKeyValue('foo', bar),
ContainKeyValue('hello', world))
def __init__(self, wantkey, wantvalue):
self.wantkey = wantkey
self.wantvalue = wantvalue
def __eq__(self, other):
return other[self.wantkey] == self.wantvalue
except (KeyError, TypeError):
return False
def __ne__(self, other):
return other[self.wantkey] != self.wantvalue
except (KeyError, TypeError):
return True
def __repr__(self):
return "<ContainKeyValue: key " + str(self.wantkey) + \
" and value " + str(self.wantvalue) + ">"
def patch_exists(patched_path, result):
"""Selectively patch os.path.exists() so that if it's called with
patched_path, return result. Calls with any other path are passed
through to the real os.path.exists() function.
Either import and use as a decorator / context manager, or use the
nova.TestCase.patch_exists() static method as a context manager.
Currently it is *not* recommended to use this if any of the
following apply:
- You want to patch via decorator *and* make assertions about how the
mock is called (since using it in the decorator form will not make
the mock available to your code).
- You want the result of the patched exists() call to be determined
programmatically (e.g. by matching substrings of patched_path).
- You expect exists() to be called multiple times on the same path
and return different values each time.
Additionally within unit tests which only test a very limited code
path, it may be possible to ensure that the code path only invokes
exists() once, in which case it's slightly overkill to do
selective patching based on the path. In this case something like
like this may be more appropriate:
@mock.patch('os.path.exists', return_value=True)
def test_my_code(self, mock_exists):
real_exists = os.path.exists
def fake_exists(path):
if path == patched_path:
return result
return real_exists(path)
with mock.patch.object(os.path, "exists") as mock_exists:
mock_exists.side_effect = fake_exists
yield mock_exists
def patch_open(patched_path, read_data):
"""Selectively patch open() so that if it's called with patched_path,
return a mock which makes it look like the file contains
read_data. Calls with any other path are passed through to the
real open() function.
Either import and use as a decorator, or use the
nova.TestCase.patch_open() static method as a context manager.
Currently it is *not* recommended to use this if any of the
following apply:
- The code under test will attempt to write to patched_path.
- You want to patch via decorator *and* make assertions about how the
mock is called (since using it in the decorator form will not make
the mock available to your code).
- You want the faked file contents to be determined
programmatically (e.g. by matching substrings of patched_path).
- You expect open() to be called multiple times on the same path
and return different file contents each time.
Additionally within unit tests which only test a very limited code
path, it may be possible to ensure that the code path only invokes
open() once, in which case it's slightly overkill to do
selective patching based on the path. In this case something like
like this may be more appropriate:
def test_my_code(self, mock_open):
real_open =
m = mock.mock_open(read_data=read_data)
def selective_fake_open(path, *args, **kwargs):
if path == patched_path:
return m(patched_path)
return real_open(path, *args, **kwargs)
with mock.patch('') as mock_open:
mock_open.side_effect = selective_fake_open
yield m