deb-heat/heat/tests/test_autoscaling_update_policy.py

791 lines
32 KiB
Python

#
# 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.
import copy
import json
import mox
from oslo.config import cfg
from testtools.matchers import MatchesRegex
from heat.common import exception
from heat.common import template_format
from heat.engine import clients
from heat.engine.notification import stack as notification
from heat.engine import parser
from heat.engine.resources import instance
from heat.engine.resources import loadbalancer as lb
from heat.engine.resources import wait_condition as wc
from heat.engine import stack_user
from heat.tests.common import HeatTestCase
from heat.tests import fakes
from heat.tests import utils
from heat.tests.v1_1 import fakes as fakes11
asg_tmpl_without_updt_policy = '''
{
"AWSTemplateFormatVersion" : "2010-09-09",
"Description" : "Template to create autoscaling group.",
"Parameters" : {},
"Resources" : {
"WebServerGroup" : {
"Type" : "AWS::AutoScaling::AutoScalingGroup",
"Properties" : {
"AvailabilityZones" : ["nova"],
"LaunchConfigurationName" : { "Ref" : "LaunchConfig" },
"MinSize" : "10",
"MaxSize" : "20",
"LoadBalancerNames" : [ { "Ref" : "ElasticLoadBalancer" } ]
}
},
"ElasticLoadBalancer" : {
"Type" : "AWS::ElasticLoadBalancing::LoadBalancer",
"Properties" : {
"AvailabilityZones" : ["nova"],
"Listeners" : [ {
"LoadBalancerPort" : "80",
"InstancePort" : "80",
"Protocol" : "HTTP"
}]
}
},
"LaunchConfig" : {
"Type" : "AWS::AutoScaling::LaunchConfiguration",
"Properties": {
"ImageId" : "F20-x86_64-cfntools",
"InstanceType" : "m1.medium",
"KeyName" : "test",
"SecurityGroups" : [ "sg-1" ],
"UserData" : "jsconfig data"
}
}
}
}
'''
asg_tmpl_with_bad_updt_policy = '''
{
"AWSTemplateFormatVersion" : "2010-09-09",
"Description" : "Template to create autoscaling group.",
"Parameters" : {},
"Resources" : {
"WebServerGroup" : {
"UpdatePolicy": {
"foo": {
}
},
"Type" : "AWS::AutoScaling::AutoScalingGroup",
"Properties" : {
"AvailabilityZones" : ["nova"],
"LaunchConfigurationName" : { "Ref" : "LaunchConfig" },
"MinSize" : "10",
"MaxSize" : "20"
}
},
"LaunchConfig" : {
"Type" : "AWS::AutoScaling::LaunchConfiguration",
"Properties": {
"ImageId" : "F20-x86_64-cfntools",
"InstanceType" : "m1.medium",
"KeyName" : "test",
"SecurityGroups" : [ "sg-1" ],
"UserData" : "jsconfig data"
}
}
}
}
'''
asg_tmpl_with_default_updt_policy = '''
{
"AWSTemplateFormatVersion" : "2010-09-09",
"Description" : "Template to create autoscaling group.",
"Parameters" : {},
"Resources" : {
"WebServerGroup" : {
"UpdatePolicy" : {
"AutoScalingRollingUpdate" : {
}
},
"Type" : "AWS::AutoScaling::AutoScalingGroup",
"Properties" : {
"AvailabilityZones" : ["nova"],
"LaunchConfigurationName" : { "Ref" : "LaunchConfig" },
"MinSize" : "10",
"MaxSize" : "20",
"LoadBalancerNames" : [ { "Ref" : "ElasticLoadBalancer" } ]
}
},
"ElasticLoadBalancer" : {
"Type" : "AWS::ElasticLoadBalancing::LoadBalancer",
"Properties" : {
"AvailabilityZones" : ["nova"],
"Listeners" : [ {
"LoadBalancerPort" : "80",
"InstancePort" : "80",
"Protocol" : "HTTP"
}]
}
},
"LaunchConfig" : {
"Type" : "AWS::AutoScaling::LaunchConfiguration",
"Properties": {
"ImageId" : "F20-x86_64-cfntools",
"InstanceType" : "m1.medium",
"KeyName" : "test",
"SecurityGroups" : [ "sg-1" ],
"UserData" : "jsconfig data"
}
}
}
}
'''
asg_tmpl_with_updt_policy = '''
{
"AWSTemplateFormatVersion" : "2010-09-09",
"Description" : "Template to create autoscaling group.",
"Parameters" : {},
"Resources" : {
"WebServerGroup" : {
"UpdatePolicy" : {
"AutoScalingRollingUpdate" : {
"MinInstancesInService" : "1",
"MaxBatchSize" : "2",
"PauseTime" : "PT1S"
}
},
"Type" : "AWS::AutoScaling::AutoScalingGroup",
"Properties" : {
"AvailabilityZones" : ["nova"],
"LaunchConfigurationName" : { "Ref" : "LaunchConfig" },
"MinSize" : "10",
"MaxSize" : "20",
"LoadBalancerNames" : [ { "Ref" : "ElasticLoadBalancer" } ]
}
},
"ElasticLoadBalancer" : {
"Type" : "AWS::ElasticLoadBalancing::LoadBalancer",
"Properties" : {
"AvailabilityZones" : ["nova"],
"Listeners" : [ {
"LoadBalancerPort" : "80",
"InstancePort" : "80",
"Protocol" : "HTTP"
}]
}
},
"LaunchConfig" : {
"Type" : "AWS::AutoScaling::LaunchConfiguration",
"Properties": {
"ImageId" : "F20-x86_64-cfntools",
"InstanceType" : "m1.medium",
"KeyName" : "test",
"SecurityGroups" : [ "sg-1" ],
"UserData" : "jsconfig data"
}
}
}
}
'''
class AutoScalingGroupTest(HeatTestCase):
def setUp(self):
super(AutoScalingGroupTest, self).setUp()
self.fc = fakes11.FakeClient()
self.fkc = fakes.FakeKeystoneClient(username='test_stack.CfnLBUser')
cfg.CONF.set_default('heat_waitcondition_server_url',
'http://127.0.0.1:8000/v1/waitcondition')
utils.setup_dummy_db()
def _stub_validate(self):
self.m.StubOutWithMock(parser.Stack, 'validate')
parser.Stack.validate().MultipleTimes()
def _stub_lb_create(self):
self.m.StubOutWithMock(stack_user.StackUser, 'keystone')
stack_user.StackUser.keystone().MultipleTimes().AndReturn(self.fkc)
self.m.StubOutWithMock(wc.WaitConditionHandle, 'get_status')
wc.WaitConditionHandle.get_status().AndReturn(['SUCCESS'])
def _stub_lb_reload(self, num=1, setup=True):
if setup:
self.m.StubOutWithMock(lb.LoadBalancer, 'handle_update')
for i in range(num):
lb.LoadBalancer.handle_update(
mox.IgnoreArg(), mox.IgnoreArg(),
mox.IgnoreArg()).AndReturn(None)
def _stub_grp_create(self, capacity=0, setup_lb=True):
"""
Expect creation of instances to capacity. By default, expect creation
of load balancer unless specified.
"""
self._stub_validate()
self.m.StubOutWithMock(clients.OpenStackClients, 'nova')
self.m.StubOutWithMock(instance.Instance, 'handle_create')
self.m.StubOutWithMock(instance.Instance, 'check_create_complete')
self.m.StubOutWithMock(notification, 'send')
notification.send(mox.IgnoreArg()).MultipleTimes().AndReturn(None)
cookie = object()
clients.OpenStackClients.nova().MultipleTimes().AndReturn(self.fc)
# for load balancer setup
if setup_lb:
self._stub_lb_create()
self._stub_lb_reload()
instance.Instance.handle_create().AndReturn(cookie)
instance.Instance.check_create_complete(cookie).AndReturn(True)
# for each instance in group
for i in range(capacity):
instance.Instance.handle_create().AndReturn(cookie)
instance.Instance.check_create_complete(cookie).AndReturn(True)
def _stub_grp_replace(self,
num_creates_expected_on_updt=0,
num_deletes_expected_on_updt=0,
num_reloads_expected_on_updt=0):
"""
Expect replacement of the capacity by batch size
"""
# for load balancer setup
self._stub_lb_reload(num_reloads_expected_on_updt)
self.m.StubOutWithMock(notification, 'send')
notification.send(mox.IgnoreArg()).MultipleTimes().AndReturn(None)
# for instances in the group
self.m.StubOutWithMock(clients.OpenStackClients, 'nova')
self.m.StubOutWithMock(instance.Instance, 'handle_create')
self.m.StubOutWithMock(instance.Instance, 'check_create_complete')
self.m.StubOutWithMock(instance.Instance, 'destroy')
if num_reloads_expected_on_updt > 1:
clients.OpenStackClients.nova().MultipleTimes().AndReturn(self.fc)
cookie = object()
for i in range(num_creates_expected_on_updt):
instance.Instance.handle_create().AndReturn(cookie)
instance.Instance.check_create_complete(cookie).AndReturn(True)
for i in range(num_deletes_expected_on_updt):
instance.Instance.destroy().AndReturn(None)
def _stub_grp_update(self,
num_creates_expected_on_updt=0,
num_deletes_expected_on_updt=0,
num_reloads_expected_on_updt=0):
"""
Expect update of the instances
"""
self.m.StubOutWithMock(instance.Instance, 'nova')
instance.Instance.nova().MultipleTimes().AndReturn(self.fc)
def activate_status(server):
server.status = 'VERIFY_RESIZE'
return_server = self.fc.servers.list()[1]
return_server.id = 1234
return_server.get = activate_status.__get__(return_server)
self.m.StubOutWithMock(self.fc.servers, 'get')
self.m.StubOutWithMock(self.fc.client, 'post_servers_1234_action')
self.fc.servers.get(mox.IgnoreArg()).\
MultipleTimes().AndReturn(return_server)
self.fc.client.post_servers_1234_action(
body={'resize': {'flavorRef': 3}}).\
MultipleTimes().AndReturn((202, None))
self.fc.client.post_servers_1234_action(
body={'confirmResize': None}).\
MultipleTimes().AndReturn((202, None))
self._stub_grp_replace(num_creates_expected_on_updt,
num_deletes_expected_on_updt,
num_reloads_expected_on_updt)
def get_launch_conf_name(self, stack, ig_name):
return stack[ig_name].properties['LaunchConfigurationName']
def test_parse_without_update_policy(self):
tmpl = template_format.parse(asg_tmpl_without_updt_policy)
stack = utils.parse_stack(tmpl)
stack.validate()
grp = stack['WebServerGroup']
self.assertFalse(grp.update_policy['AutoScalingRollingUpdate'])
def test_parse_with_update_policy(self):
tmpl = template_format.parse(asg_tmpl_with_updt_policy)
stack = utils.parse_stack(tmpl)
stack.validate()
tmpl_grp = tmpl['Resources']['WebServerGroup']
tmpl_policy = tmpl_grp['UpdatePolicy']['AutoScalingRollingUpdate']
tmpl_batch_sz = int(tmpl_policy['MaxBatchSize'])
grp = stack['WebServerGroup']
self.assertTrue(grp.update_policy)
self.assertEqual(1, len(grp.update_policy))
self.assertIn('AutoScalingRollingUpdate', grp.update_policy)
policy = grp.update_policy['AutoScalingRollingUpdate']
self.assertTrue(policy and len(policy) > 0)
self.assertEqual(1, int(policy['MinInstancesInService']))
self.assertEqual(tmpl_batch_sz, int(policy['MaxBatchSize']))
self.assertEqual('PT1S', policy['PauseTime'])
def test_parse_with_default_update_policy(self):
tmpl = template_format.parse(asg_tmpl_with_default_updt_policy)
stack = utils.parse_stack(tmpl)
stack.validate()
grp = stack['WebServerGroup']
self.assertTrue(grp.update_policy)
self.assertEqual(1, len(grp.update_policy))
self.assertIn('AutoScalingRollingUpdate', grp.update_policy)
policy = grp.update_policy['AutoScalingRollingUpdate']
self.assertTrue(policy and len(policy) > 0)
self.assertEqual(0, int(policy['MinInstancesInService']))
self.assertEqual(1, int(policy['MaxBatchSize']))
self.assertEqual('PT0S', policy['PauseTime'])
def test_parse_with_bad_update_policy(self):
tmpl = template_format.parse(asg_tmpl_with_bad_updt_policy)
stack = utils.parse_stack(tmpl)
self.assertRaises(exception.StackValidationFailed, stack.validate)
def test_parse_with_bad_pausetime_in_update_policy(self):
tmpl = template_format.parse(asg_tmpl_with_default_updt_policy)
group = tmpl['Resources']['WebServerGroup']
policy = group['UpdatePolicy']['AutoScalingRollingUpdate']
policy['PauseTime'] = 'P1YT1H'
stack = utils.parse_stack(tmpl)
self.assertRaises(exception.StackValidationFailed, stack.validate)
def validate_update_policy_diff(self, current, updated):
# load current stack
current_tmpl = template_format.parse(current)
current_stack = utils.parse_stack(current_tmpl)
# get the json snippet for the current InstanceGroup resource
current_grp = current_stack['WebServerGroup']
current_snippets = dict((n, r.parsed_template())
for n, r in current_stack.items())
current_grp_json = current_snippets[current_grp.name]
# load the updated stack
updated_tmpl = template_format.parse(updated)
updated_stack = utils.parse_stack(updated_tmpl)
# get the updated json snippet for the InstanceGroup resource in the
# context of the current stack
updated_grp = updated_stack['WebServerGroup']
updated_grp_json = current_stack.resolve_runtime_data(updated_grp.t)
# identify the template difference
tmpl_diff = updated_grp.update_template_diff(
updated_grp_json, current_grp_json)
updated_policy = (updated_grp.t['UpdatePolicy']
if 'UpdatePolicy' in updated_grp.t else None)
expected = {u'UpdatePolicy': updated_policy}
self.assertEqual(expected, tmpl_diff)
def test_update_policy_added(self):
self.validate_update_policy_diff(asg_tmpl_without_updt_policy,
asg_tmpl_with_updt_policy)
def test_update_policy_updated(self):
updt_template = json.loads(asg_tmpl_with_updt_policy)
grp = updt_template['Resources']['WebServerGroup']
policy = grp['UpdatePolicy']['AutoScalingRollingUpdate']
policy['MinInstancesInService'] = '2'
policy['MaxBatchSize'] = '4'
policy['PauseTime'] = 'PT1M30S'
self.validate_update_policy_diff(asg_tmpl_with_updt_policy,
json.dumps(updt_template))
def test_update_policy_removed(self):
self.validate_update_policy_diff(asg_tmpl_with_updt_policy,
asg_tmpl_without_updt_policy)
def update_autoscaling_group(self, init_template, updt_template,
num_updates_expected_on_updt,
num_creates_expected_on_updt,
num_deletes_expected_on_updt,
num_reloads_expected_on_updt,
update_replace):
# setup stack from the initial template
tmpl = template_format.parse(init_template)
stack = utils.parse_stack(tmpl)
stack.validate()
# test stack create
size = int(stack['WebServerGroup'].properties['MinSize'])
self._stub_grp_create(size)
self.m.ReplayAll()
stack.create()
self.m.VerifyAll()
self.assertEqual(('CREATE', 'COMPLETE'), stack.state)
# test that update policy is loaded
current_grp = stack['WebServerGroup']
self.assertTrue('AutoScalingRollingUpdate'
in current_grp.update_policy)
current_policy = current_grp.update_policy['AutoScalingRollingUpdate']
self.assertTrue(current_policy)
self.assertTrue(len(current_policy) > 0)
init_updt_policy = tmpl['Resources']['WebServerGroup']['UpdatePolicy']
init_roll_updt = init_updt_policy['AutoScalingRollingUpdate']
init_batch_sz = int(init_roll_updt['MaxBatchSize'])
self.assertEqual(init_batch_sz, int(current_policy['MaxBatchSize']))
# test that physical resource name of launch configuration is used
conf = stack['LaunchConfig']
conf_name_pattern = '%s-LaunchConfig-[a-zA-Z0-9]+$' % stack.name
self.assertThat(conf.FnGetRefId(), MatchesRegex(conf_name_pattern))
# get launch conf name here to compare result after update
conf_name = self.get_launch_conf_name(stack, 'WebServerGroup')
# test the number of instances created
nested = stack['WebServerGroup'].nested()
self.assertEqual(size, len(nested.resources))
# clean up for next test
self.m.UnsetStubs()
# saves info from initial list of instances for comparison later
init_instances = current_grp.get_instances()
init_names = current_grp.get_instance_names()
init_images = [(i.name, i.t['Properties']['ImageId'])
for i in init_instances]
init_flavors = [(i.name, i.t['Properties']['InstanceType'])
for i in init_instances]
# test stack update
updated_tmpl = template_format.parse(updt_template)
updated_stack = utils.parse_stack(updated_tmpl)
new_grp_tmpl = updated_tmpl['Resources']['WebServerGroup']
new_updt_pol = new_grp_tmpl['UpdatePolicy']['AutoScalingRollingUpdate']
new_batch_sz = int(new_updt_pol['MaxBatchSize'])
self.assertNotEqual(new_batch_sz, init_batch_sz)
self._stub_validate()
if update_replace:
self._stub_grp_replace(size, size, num_reloads_expected_on_updt)
else:
self._stub_grp_update(num_creates_expected_on_updt,
num_deletes_expected_on_updt,
num_reloads_expected_on_updt)
self.stub_wallclock()
self.m.ReplayAll()
stack.update(updated_stack)
self.m.VerifyAll()
self.assertEqual(('UPDATE', 'COMPLETE'), stack.state)
# test that the update policy is updated
updated_grp = stack['WebServerGroup']
updt_instances = updated_grp.get_instances()
self.assertTrue('AutoScalingRollingUpdate'
in updated_grp.update_policy)
updated_policy = updated_grp.update_policy['AutoScalingRollingUpdate']
self.assertTrue(updated_policy)
self.assertTrue(len(updated_policy) > 0)
self.assertEqual(new_batch_sz, int(updated_policy['MaxBatchSize']))
# test that the launch configuration is replaced
updated_conf_name = self.get_launch_conf_name(stack, 'WebServerGroup')
self.assertNotEqual(conf_name, updated_conf_name)
# test that the group size are the same
updt_instances = updated_grp.get_instances()
updt_names = updated_grp.get_instance_names()
self.assertEqual(len(init_names), len(updt_names))
# test that appropriate number of instance names are the same
matched_names = set(updt_names) & set(init_names)
self.assertEqual(num_updates_expected_on_updt, len(matched_names))
# test that the appropriate number of new instances are created
self.assertEqual(num_creates_expected_on_updt,
len(set(updt_names) - set(init_names)))
# test that the appropriate number of instances are deleted
self.assertEqual(num_deletes_expected_on_updt,
len(set(init_names) - set(updt_names)))
# test that the older instances are the ones being deleted
if num_deletes_expected_on_updt > 0:
deletes_expected = init_names[:num_deletes_expected_on_updt]
self.assertNotIn(deletes_expected, updt_names)
# test if instances are updated
if update_replace:
# test that the image id is changed for all instances
updt_images = [(i.name, i.t['Properties']['ImageId'])
for i in updt_instances]
self.assertEqual(0, len(set(updt_images) & set(init_images)))
else:
# test that instance type is changed for all instances
updt_flavors = [(i.name, i.t['Properties']['InstanceType'])
for i in updt_instances]
self.assertEqual(0, len(set(updt_flavors) & set(init_flavors)))
def test_autoscaling_group_update_replace(self):
"""
Test simple update replace with no conflict in batch size and
minimum instances in service.
"""
updt_template = json.loads(asg_tmpl_with_updt_policy)
grp = updt_template['Resources']['WebServerGroup']
policy = grp['UpdatePolicy']['AutoScalingRollingUpdate']
policy['MinInstancesInService'] = '1'
policy['MaxBatchSize'] = '3'
config = updt_template['Resources']['LaunchConfig']
config['Properties']['ImageId'] = 'F17-x86_64-cfntools'
self.update_autoscaling_group(asg_tmpl_with_updt_policy,
json.dumps(updt_template),
num_updates_expected_on_updt=10,
num_creates_expected_on_updt=0,
num_deletes_expected_on_updt=0,
num_reloads_expected_on_updt=9,
update_replace=True)
def test_autoscaling_group_update_replace_with_adjusted_capacity(self):
"""
Test update replace with capacity adjustment due to conflict in
batch size and minimum instances in service.
"""
updt_template = json.loads(asg_tmpl_with_updt_policy)
grp = updt_template['Resources']['WebServerGroup']
policy = grp['UpdatePolicy']['AutoScalingRollingUpdate']
policy['MinInstancesInService'] = '8'
policy['MaxBatchSize'] = '4'
config = updt_template['Resources']['LaunchConfig']
config['Properties']['ImageId'] = 'F17-x86_64-cfntools'
self.update_autoscaling_group(asg_tmpl_with_updt_policy,
json.dumps(updt_template),
num_updates_expected_on_updt=8,
num_creates_expected_on_updt=2,
num_deletes_expected_on_updt=2,
num_reloads_expected_on_updt=7,
update_replace=True)
def test_autoscaling_group_update_replace_huge_batch_size(self):
"""
Test update replace with a huge batch size.
"""
updt_template = json.loads(asg_tmpl_with_updt_policy)
group = updt_template['Resources']['WebServerGroup']
policy = group['UpdatePolicy']['AutoScalingRollingUpdate']
policy['MinInstancesInService'] = '0'
policy['MaxBatchSize'] = '20'
config = updt_template['Resources']['LaunchConfig']
config['Properties']['ImageId'] = 'F17-x86_64-cfntools'
self.update_autoscaling_group(asg_tmpl_with_updt_policy,
json.dumps(updt_template),
num_updates_expected_on_updt=10,
num_creates_expected_on_updt=0,
num_deletes_expected_on_updt=0,
num_reloads_expected_on_updt=3,
update_replace=True)
def test_autoscaling_group_update_replace_huge_min_in_service(self):
"""
Test update replace with a huge number of minimum instances in service.
"""
updt_template = json.loads(asg_tmpl_with_updt_policy)
group = updt_template['Resources']['WebServerGroup']
policy = group['UpdatePolicy']['AutoScalingRollingUpdate']
policy['MinInstancesInService'] = '20'
policy['MaxBatchSize'] = '1'
policy['PauseTime'] = 'PT0S'
config = updt_template['Resources']['LaunchConfig']
config['Properties']['ImageId'] = 'F17-x86_64-cfntools'
self.update_autoscaling_group(asg_tmpl_with_updt_policy,
json.dumps(updt_template),
num_updates_expected_on_updt=9,
num_creates_expected_on_updt=1,
num_deletes_expected_on_updt=1,
num_reloads_expected_on_updt=12,
update_replace=True)
def test_autoscaling_group_update_no_replace(self):
"""
Test simple update only and no replace (i.e. updated instance flavor
in Launch Configuration) with no conflict in batch size and
minimum instances in service.
"""
updt_template = json.loads(copy.deepcopy(asg_tmpl_with_updt_policy))
group = updt_template['Resources']['WebServerGroup']
policy = group['UpdatePolicy']['AutoScalingRollingUpdate']
policy['MinInstancesInService'] = '1'
policy['MaxBatchSize'] = '3'
policy['PauseTime'] = 'PT0S'
config = updt_template['Resources']['LaunchConfig']
config['Properties']['InstanceType'] = 'm1.large'
self.update_autoscaling_group(asg_tmpl_with_updt_policy,
json.dumps(updt_template),
num_updates_expected_on_updt=10,
num_creates_expected_on_updt=0,
num_deletes_expected_on_updt=0,
num_reloads_expected_on_updt=6,
update_replace=False)
def test_instance_group_update_no_replace_with_adjusted_capacity(self):
"""
Test update only and no replace (i.e. updated instance flavor in
Launch Configuration) with capacity adjustment due to conflict in
batch size and minimum instances in service.
"""
updt_template = json.loads(copy.deepcopy(asg_tmpl_with_updt_policy))
group = updt_template['Resources']['WebServerGroup']
policy = group['UpdatePolicy']['AutoScalingRollingUpdate']
policy['MinInstancesInService'] = '8'
policy['MaxBatchSize'] = '4'
policy['PauseTime'] = 'PT0S'
config = updt_template['Resources']['LaunchConfig']
config['Properties']['InstanceType'] = 'm1.large'
self.update_autoscaling_group(asg_tmpl_with_updt_policy,
json.dumps(updt_template),
num_updates_expected_on_updt=8,
num_creates_expected_on_updt=2,
num_deletes_expected_on_updt=2,
num_reloads_expected_on_updt=5,
update_replace=False)
def test_autoscaling_group_update_policy_removed(self):
# setup stack from the initial template
tmpl = template_format.parse(asg_tmpl_with_updt_policy)
stack = utils.parse_stack(tmpl)
stack.validate()
# test stack create
size = int(stack['WebServerGroup'].properties['MinSize'])
self._stub_grp_create(size)
self.m.ReplayAll()
stack.create()
self.m.VerifyAll()
self.assertEqual(('CREATE', 'COMPLETE'), stack.state)
# test that update policy is loaded
current_grp = stack['WebServerGroup']
self.assertIn('AutoScalingRollingUpdate', current_grp.update_policy)
current_policy = current_grp.update_policy['AutoScalingRollingUpdate']
self.assertTrue(current_policy)
self.assertTrue(len(current_policy) > 0)
init_updt_policy = tmpl['Resources']['WebServerGroup']['UpdatePolicy']
init_roll_updt = init_updt_policy['AutoScalingRollingUpdate']
init_batch_sz = int(init_roll_updt['MaxBatchSize'])
self.assertEqual(init_batch_sz, int(current_policy['MaxBatchSize']))
# test that physical resource name of launch configuration is used
conf = stack['LaunchConfig']
conf_name_pattern = '%s-LaunchConfig-[a-zA-Z0-9]+$' % stack.name
self.assertThat(conf.FnGetRefId(), MatchesRegex(conf_name_pattern))
# test the number of instances created
nested = stack['WebServerGroup'].nested()
self.assertEqual(size, len(nested.resources))
# clean up for next test
self.m.UnsetStubs()
# test stack update
updated_tmpl = template_format.parse(asg_tmpl_without_updt_policy)
updated_stack = utils.parse_stack(updated_tmpl)
self._stub_grp_replace(num_creates_expected_on_updt=0,
num_deletes_expected_on_updt=0,
num_reloads_expected_on_updt=1)
self.m.ReplayAll()
stack.update(updated_stack)
self.m.VerifyAll()
self.assertEqual(('UPDATE', 'COMPLETE'), stack.state)
# test that update policy is removed
updated_grp = stack['WebServerGroup']
self.assertFalse(updated_grp.update_policy['AutoScalingRollingUpdate'])
def test_autoscaling_group_update_policy_check_timeout(self):
# setup stack from the initial template
tmpl = template_format.parse(asg_tmpl_with_updt_policy)
stack = utils.parse_stack(tmpl)
# test stack create
size = int(stack['WebServerGroup'].properties['MinSize'])
self._stub_grp_create(size)
self.m.ReplayAll()
stack.create()
self.m.VerifyAll()
self.assertEqual(('CREATE', 'COMPLETE'), stack.state)
# test that update policy is loaded
current_grp = stack['WebServerGroup']
self.assertIn('AutoScalingRollingUpdate', current_grp.update_policy)
current_policy = current_grp.update_policy['AutoScalingRollingUpdate']
self.assertTrue(current_policy)
self.assertTrue(len(current_policy) > 0)
init_updt_policy = tmpl['Resources']['WebServerGroup']['UpdatePolicy']
init_roll_updt = init_updt_policy['AutoScalingRollingUpdate']
init_batch_sz = int(init_roll_updt['MaxBatchSize'])
self.assertEqual(init_batch_sz, int(current_policy['MaxBatchSize']))
# test the number of instances created
nested = stack['WebServerGroup'].nested()
self.assertEqual(size, len(nested.resources))
# clean up for next test
self.m.UnsetStubs()
# modify the pause time and test for error
new_pause_time = 'PT30M'
updt_template = json.loads(copy.deepcopy(asg_tmpl_with_updt_policy))
group = updt_template['Resources']['WebServerGroup']
policy = group['UpdatePolicy']['AutoScalingRollingUpdate']
policy['PauseTime'] = new_pause_time
config = updt_template['Resources']['LaunchConfig']
config['Properties']['ImageId'] = 'F17-x86_64-cfntools'
updated_tmpl = template_format.parse(json.dumps(updt_template))
updated_stack = utils.parse_stack(updated_tmpl)
self._stub_grp_replace(num_creates_expected_on_updt=0,
num_deletes_expected_on_updt=0,
num_reloads_expected_on_updt=1)
self.m.ReplayAll()
stack.update(updated_stack)
self.m.VerifyAll()
self.assertEqual(('UPDATE', 'FAILED'), stack.state)
# test that the update policy is updated
updated_grp = stack['WebServerGroup']
self.assertIn('AutoScalingRollingUpdate', updated_grp.update_policy)
updated_policy = updated_grp.update_policy['AutoScalingRollingUpdate']
self.assertTrue(updated_policy)
self.assertTrue(len(updated_policy) > 0)
self.assertEqual(new_pause_time, updated_policy['PauseTime'])
# test that error message match
expected_error_message = ('The current UpdatePolicy will result '
'in stack update timeout.')
self.assertIn(expected_error_message, stack.status_reason)