
Add return-request-id-to-caller function to resources and resource managers in the following files. The methods in the resource class and resource manager return a wrapper class that has 'request_ids' property. The caller can get request ids of the callee via the property. * novaclient/v2/agents.py * novaclient/v2/aggregates.py * novaclient/v2/availability_zones.py * novaclient/v2/certs.py * novaclient/v2/cloudpipe.py * novaclient/v2/fixed_ips.py * novaclient/v2/flavor_access.py * novaclient/v2/flavors.py * novaclient/v2/floating_ip_dns.py * novaclient/v2/floating_ip_pools.py * novaclient/v2/floating_ips_bulk.py * novaclient/v2/floating_ips.py * novaclient/v2/fping.py * novaclient/v2/hosts.py * novaclient/v2/hypervisors.py * novaclient/v2/images.py Co-authored-by: Ankit Agrawal <ankit11.agrawal@nttdata.com> Change-Id: Ic624d532f1a468dd45105bbb5aa6d8a730984338 Implements: blueprint return-request-id-to-caller
188 lines
6.6 KiB
Python
188 lines
6.6 KiB
Python
# Copyright 2012 OpenStack Foundation
|
|
# 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
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# 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.
|
|
|
|
from novaclient.tests.unit.fixture_data import client
|
|
from novaclient.tests.unit.fixture_data import hypervisors as data
|
|
from novaclient.tests.unit import utils
|
|
from novaclient.tests.unit.v2 import fakes
|
|
|
|
|
|
class HypervisorsTest(utils.FixturedTestCase):
|
|
|
|
client_fixture_class = client.V1
|
|
data_fixture_class = data.V1
|
|
|
|
def compare_to_expected(self, expected, hyper):
|
|
for key, value in expected.items():
|
|
self.assertEqual(getattr(hyper, key), value)
|
|
|
|
def test_hypervisor_index(self):
|
|
expected = [
|
|
dict(id=1234, hypervisor_hostname='hyper1'),
|
|
dict(id=5678, hypervisor_hostname='hyper2')]
|
|
|
|
result = self.cs.hypervisors.list(False)
|
|
self.assert_request_id(result, fakes.FAKE_REQUEST_ID_LIST)
|
|
self.assert_called('GET', '/os-hypervisors')
|
|
|
|
for idx, hyper in enumerate(result):
|
|
self.compare_to_expected(expected[idx], hyper)
|
|
|
|
def test_hypervisor_detail(self):
|
|
expected = [
|
|
dict(id=1234,
|
|
service=dict(id=1, host='compute1'),
|
|
vcpus=4,
|
|
memory_mb=10 * 1024,
|
|
local_gb=250,
|
|
vcpus_used=2,
|
|
memory_mb_used=5 * 1024,
|
|
local_gb_used=125,
|
|
hypervisor_type="xen",
|
|
hypervisor_version=3,
|
|
hypervisor_hostname="hyper1",
|
|
free_ram_mb=5 * 1024,
|
|
free_disk_gb=125,
|
|
current_workload=2,
|
|
running_vms=2,
|
|
cpu_info='cpu_info',
|
|
disk_available_least=100),
|
|
dict(id=2,
|
|
service=dict(id=2, host="compute2"),
|
|
vcpus=4,
|
|
memory_mb=10 * 1024,
|
|
local_gb=250,
|
|
vcpus_used=2,
|
|
memory_mb_used=5 * 1024,
|
|
local_gb_used=125,
|
|
hypervisor_type="xen",
|
|
hypervisor_version=3,
|
|
hypervisor_hostname="hyper2",
|
|
free_ram_mb=5 * 1024,
|
|
free_disk_gb=125,
|
|
current_workload=2,
|
|
running_vms=2,
|
|
cpu_info='cpu_info',
|
|
disk_available_least=100)]
|
|
|
|
result = self.cs.hypervisors.list()
|
|
self.assert_request_id(result, fakes.FAKE_REQUEST_ID_LIST)
|
|
self.assert_called('GET', '/os-hypervisors/detail')
|
|
|
|
for idx, hyper in enumerate(result):
|
|
self.compare_to_expected(expected[idx], hyper)
|
|
|
|
def test_hypervisor_search(self):
|
|
expected = [
|
|
dict(id=1234, hypervisor_hostname='hyper1'),
|
|
dict(id=5678, hypervisor_hostname='hyper2')]
|
|
|
|
result = self.cs.hypervisors.search('hyper')
|
|
self.assert_request_id(result, fakes.FAKE_REQUEST_ID_LIST)
|
|
self.assert_called('GET', '/os-hypervisors/hyper/search')
|
|
|
|
for idx, hyper in enumerate(result):
|
|
self.compare_to_expected(expected[idx], hyper)
|
|
|
|
def test_hypervisor_servers(self):
|
|
expected = [
|
|
dict(id=1234,
|
|
hypervisor_hostname='hyper1',
|
|
servers=[
|
|
dict(name='inst1', uuid='uuid1'),
|
|
dict(name='inst2', uuid='uuid2')]),
|
|
dict(id=5678,
|
|
hypervisor_hostname='hyper2',
|
|
servers=[
|
|
dict(name='inst3', uuid='uuid3'),
|
|
dict(name='inst4', uuid='uuid4')]),
|
|
]
|
|
|
|
result = self.cs.hypervisors.search('hyper', True)
|
|
self.assert_request_id(result, fakes.FAKE_REQUEST_ID_LIST)
|
|
self.assert_called('GET', '/os-hypervisors/hyper/servers')
|
|
|
|
for idx, hyper in enumerate(result):
|
|
self.compare_to_expected(expected[idx], hyper)
|
|
|
|
def test_hypervisor_get(self):
|
|
expected = dict(
|
|
id=1234,
|
|
service=dict(id=1, host='compute1'),
|
|
vcpus=4,
|
|
memory_mb=10 * 1024,
|
|
local_gb=250,
|
|
vcpus_used=2,
|
|
memory_mb_used=5 * 1024,
|
|
local_gb_used=125,
|
|
hypervisor_type="xen",
|
|
hypervisor_version=3,
|
|
hypervisor_hostname="hyper1",
|
|
free_ram_mb=5 * 1024,
|
|
free_disk_gb=125,
|
|
current_workload=2,
|
|
running_vms=2,
|
|
cpu_info='cpu_info',
|
|
disk_available_least=100)
|
|
|
|
result = self.cs.hypervisors.get(1234)
|
|
self.assert_request_id(result, fakes.FAKE_REQUEST_ID_LIST)
|
|
self.assert_called('GET', '/os-hypervisors/1234')
|
|
|
|
self.compare_to_expected(expected, result)
|
|
|
|
def test_hypervisor_uptime(self):
|
|
expected = dict(
|
|
id=1234,
|
|
hypervisor_hostname="hyper1",
|
|
uptime="fake uptime")
|
|
|
|
result = self.cs.hypervisors.uptime(1234)
|
|
self.assert_request_id(result, fakes.FAKE_REQUEST_ID_LIST)
|
|
self.assert_called('GET', '/os-hypervisors/1234/uptime')
|
|
|
|
self.compare_to_expected(expected, result)
|
|
|
|
def test_hypervisor_statistics(self):
|
|
expected = dict(
|
|
count=2,
|
|
vcpus=8,
|
|
memory_mb=20 * 1024,
|
|
local_gb=500,
|
|
vcpus_used=4,
|
|
memory_mb_used=10 * 1024,
|
|
local_gb_used=250,
|
|
free_ram_mb=10 * 1024,
|
|
free_disk_gb=250,
|
|
current_workload=4,
|
|
running_vms=4,
|
|
disk_available_least=200,
|
|
)
|
|
|
|
result = self.cs.hypervisors.statistics()
|
|
self.assert_request_id(result, fakes.FAKE_REQUEST_ID_LIST)
|
|
self.assert_called('GET', '/os-hypervisors/statistics')
|
|
|
|
self.compare_to_expected(expected, result)
|
|
|
|
def test_hypervisor_statistics_data_model(self):
|
|
result = self.cs.hypervisor_stats.statistics()
|
|
self.assert_request_id(result, fakes.FAKE_REQUEST_ID_LIST)
|
|
self.assert_called('GET', '/os-hypervisors/statistics')
|
|
|
|
# Test for Bug #1370415, the line below used to raise AttributeError
|
|
self.assertEqual("<HypervisorStats: 2 Hypervisors>",
|
|
result.__repr__())
|