Ecs service add features (#50059)

* Support UpdateService forceNewDeployment in ecs_service module

* Fixes for review

* Add force_new_deployment option to ecs_service.py

cherrypicks changes from via/ansible
Adds tests for pull request #42518
fixes backwards compatability with boto<1.8.4

* WIP commit so I don't have to stash

* WIP commit for healthcheck grace period

* WIP commit; ecs_module handles service registries

* Fix bad check for desired_count

* Add scheduling strategy test, comment out service registry test

* Fix names in ecs_cluster role main task.

* move full test run back to the end

* Change botocore version for full test to support scheduling strategy

* fix bug with desired_count==0 in amazon/ecs_service

* Fix changed checking for scheduling strategy DAEMON in ecs_service

* Pass testS

* Fix some unhelpful comments

* Add changelog for ecs_service
This commit is contained in:
Tad Merchant 2019-02-26 22:20:19 -05:00 committed by Will Thames
parent 9474f81507
commit b538e34a32
4 changed files with 246 additions and 14 deletions

View file

@ -0,0 +1,2 @@
minor_changes:
- ecs_service - adds support for service_registries and scheduling_strategies. desired_count may now be none to support scheduling_strategies

View file

@ -132,6 +132,27 @@ options:
- Seconds to wait before health checking the freshly added/updated services. This option requires botocore >= 1.8.20.
required: false
version_added: 2.8
service_registries:
description:
- describes service disovery registries this service will register with.
required: false
version_added: 2.8
suboptions:
container_name:
description:
- container name for service disovery registration
container_port:
description:
- container port for service disovery registration
arn:
description:
- Service discovery registry ARN
scheduling_strategy:
description:
- The scheduling strategy, defaults to "REPLICA" if not given to preserve previous behavior
required: false
version_added: 2.8
choices: ["DAEMON", "REPLICA"]
extends_documentation_fragment:
- aws
- ec2
@ -387,21 +408,24 @@ class EcsServiceManager:
if (expected['load_balancers'] or []) != existing['loadBalancers']:
return False
if (expected['desired_count'] or 0) != existing['desiredCount']:
return False
# expected is params. DAEMON scheduling strategy returns desired count equal to
# number of instances running; don't check desired count if scheduling strat is daemon
if (expected['scheduling_strategy'] != 'DAEMON'):
if (expected['desired_count'] or 0) != existing['desiredCount']:
return False
return True
def create_service(self, service_name, cluster_name, task_definition, load_balancers,
desired_count, client_token, role, deployment_configuration,
placement_constraints, placement_strategy, network_configuration,
launch_type, health_check_grace_period_seconds):
placement_constraints, placement_strategy, health_check_grace_period_seconds,
network_configuration, service_registries, launch_type, scheduling_strategy):
params = dict(
cluster=cluster_name,
serviceName=service_name,
taskDefinition=task_definition,
loadBalancers=load_balancers,
desiredCount=desired_count,
clientToken=client_token,
role=role,
deploymentConfiguration=deployment_configuration,
@ -414,6 +438,14 @@ class EcsServiceManager:
params['launchType'] = launch_type
if self.health_check_setable(params) and health_check_grace_period_seconds is not None:
params['healthCheckGracePeriodSeconds'] = health_check_grace_period_seconds
if service_registries:
params['serviceRegistries'] = service_registries
# desired count is not required if scheduling strategy is daemon
if desired_count is not None:
params['desiredCount'] = desired_count
if scheduling_strategy:
params['schedulingStrategy'] = scheduling_strategy
response = self.ecs.create_service(**params)
return self.jsonize(response['service'])
@ -424,14 +456,17 @@ class EcsServiceManager:
cluster=cluster_name,
service=service_name,
taskDefinition=task_definition,
desiredCount=desired_count,
deploymentConfiguration=deployment_configuration)
if network_configuration:
params['networkConfiguration'] = network_configuration
if self.health_check_setable(params):
params['healthCheckGracePeriodSeconds'] = health_check_grace_period_seconds
if force_new_deployment:
params['forceNewDeployment'] = force_new_deployment
if health_check_grace_period_seconds is not None:
params['healthCheckGracePeriodSeconds'] = health_check_grace_period_seconds
# desired count is not required if scheduling strategy is daemon
if desired_count is not None:
params['desiredCount'] = desired_count
response = self.ecs.update_service(**params)
return self.jsonize(response['service'])
@ -484,21 +519,27 @@ def main():
deployment_configuration=dict(required=False, default={}, type='dict'),
placement_constraints=dict(required=False, default=[], type='list'),
placement_strategy=dict(required=False, default=[], type='list'),
health_check_grace_period_seconds=dict(required=False, type='int'),
network_configuration=dict(required=False, type='dict', options=dict(
subnets=dict(type='list'),
security_groups=dict(type='list'),
assign_public_ip=dict(type='bool'),
assign_public_ip=dict(type='bool')
)),
launch_type=dict(required=False, choices=['EC2', 'FARGATE']),
health_check_grace_period_seconds=dict(required=False, type='int')
service_registries=dict(required=False, type='list', default=[]),
scheduling_strategy=dict(required=False, choices=['DAEMON', 'REPLICA'])
))
module = AnsibleAWSModule(argument_spec=argument_spec,
supports_check_mode=True,
required_if=[('state', 'present', ['task_definition', 'desired_count']),
required_if=[('state', 'present', ['task_definition']),
('launch_type', 'FARGATE', ['network_configuration'])],
required_together=[['load_balancers', 'role']])
if module.params['state'] == 'present' and module.params['scheduling_strategy'] == 'REPLICA':
if module.params['desired_count'] is None:
module.fail_json(msg='state is present, scheduling_strategy is REPLICA; missing desired_count')
service_mgr = EcsServiceManager(module)
if module.params['network_configuration']:
if not service_mgr.ecs_api_handles_network_configuration():
@ -511,6 +552,7 @@ def main():
DEPLOYMENT_CONFIGURATION_TYPE_MAP)
deploymentConfiguration = snake_dict_to_camel_dict(deployment_configuration)
serviceRegistries = list(map(snake_dict_to_camel_dict, module.params['service_registries']))
try:
existing = service_mgr.describe_service(module.params['cluster'], module.params['name'])
@ -525,6 +567,9 @@ def main():
if module.params['force_new_deployment']:
if not module.botocore_at_least('1.8.4'):
module.fail_json(msg='botocore needs to be version 1.8.4 or higher to use force_new_deployment')
if module.params['health_check_grace_period_seconds']:
if not module.botocore_at_least('1.8.20'):
module.fail_json(msg='botocore needs to be version 1.8.20 or higher to use health_check_grace_period_seconds')
if module.params['state'] == 'present':
@ -557,8 +602,23 @@ def main():
loadBalancer['containerPort'] = int(loadBalancer['containerPort'])
if update:
# check various parameters and boto versions and give a helpful erro in boto is not new enough for feature
if module.params['scheduling_strategy']:
if not module.botocore_at_least('1.10.37'):
module.fail_json(msg='botocore needs to be version 1.10.37 or higher to use scheduling_strategy')
elif (existing['schedulingStrategy']) != module.params['scheduling_strategy']:
module.fail_json(msg="It is not possible to update the scheduling strategy of an existing service")
if module.params['service_registries']:
if not module.botocore_at_least('1.9.15'):
module.fail_json(msg='botocore needs to be version 1.9.15 or higher to use service_registries')
elif (existing['serviceRegistries'] or []) != serviceRegistries:
module.fail_json(msg="It is not possible to update the service registries of an existing service")
if (existing['loadBalancers'] or []) != loadBalancers:
module.fail_json(msg="It is not possible to update the load balancers of an existing service")
# update required
response = service_mgr.update_service(module.params['name'],
module.params['cluster'],
@ -568,6 +628,7 @@ def main():
network_configuration,
module.params['health_check_grace_period_seconds'],
module.params['force_new_deployment'])
else:
try:
response = service_mgr.create_service(module.params['name'],
@ -580,9 +641,11 @@ def main():
deploymentConfiguration,
module.params['placement_constraints'],
module.params['placement_strategy'],
module.params['health_check_grace_period_seconds'],
network_configuration,
serviceRegistries,
module.params['launch_type'],
module.params['health_check_grace_period_seconds']
module.params['scheduling_strategy']
)
except botocore.exceptions.ClientError as e:
module.fail_json_aws(e, msg="Couldn't create service")

View file

@ -479,6 +479,95 @@
- "update_ecs_service_with_vpc.service.networkConfiguration.awsvpcConfiguration.subnets|length == 2"
- "update_ecs_service_with_vpc.service.networkConfiguration.awsvpcConfiguration.securityGroups|length == 1"
- name: create ecs_service using health_check_grace_period_seconds
ecs_service:
name: "{{ ecs_service_name }}-mft"
cluster: "{{ ecs_cluster_name }}"
load_balancers:
- targetGroupArn: "{{ elb_target_group_instance.target_group_arn }}"
containerName: "{{ ecs_task_name }}"
containerPort: "{{ ecs_task_container_port }}"
task_definition: "{{ ecs_task_name }}:{{ ecs_task_definition.taskdefinition.revision }}"
scheduling_strategy: "REPLICA"
health_check_grace_period_seconds: 10
desired_count: 1
state: present
<<: *aws_connection_info
register: ecs_service_creation_hcgp
- name: health_check_grace_period_seconds sets HealthChecGracePeriodSeconds
assert:
that:
- ecs_service_creation_hcgp.changed
- "{{ecs_service_creation_hcgp.service.healthCheckGracePeriodSeconds}} == 10"
- name: update ecs_service using health_check_grace_period_seconds
ecs_service:
name: "{{ ecs_service_name }}-mft"
cluster: "{{ ecs_cluster_name }}"
load_balancers:
- targetGroupArn: "{{ elb_target_group_instance.target_group_arn }}"
containerName: "{{ ecs_task_name }}"
containerPort: "{{ ecs_task_container_port }}"
task_definition: "{{ ecs_task_name }}:{{ ecs_task_definition.taskdefinition.revision }}"
desired_count: 1
health_check_grace_period_seconds: 30
state: present
<<: *aws_connection_info
register: ecs_service_creation_hcgp2
ignore_errors: no
- name: check that module returns success
assert:
that:
- ecs_service_creation_hcgp2.changed
- "{{ecs_service_creation_hcgp2.service.healthCheckGracePeriodSeconds}} == 30"
# until ansible supports service registries, this test can't run.
# - name: update ecs_service using service_registries
# ecs_service:
# name: "{{ ecs_service_name }}-service-registries"
# cluster: "{{ ecs_cluster_name }}"
# load_balancers:
# - targetGroupArn: "{{ elb_target_group_instance.target_group_arn }}"
# containerName: "{{ ecs_task_name }}"
# containerPort: "{{ ecs_task_container_port }}"
# service_registries:
# - containerName: "{{ ecs_task_name }}"
# containerPort: "{{ ecs_task_container_port }}"
# ### TODO: Figure out how to get a service registry ARN without a service registry module.
# registryArn: "{{ ecs_task_service_registry_arn }}"
# task_definition: "{{ ecs_task_name }}:{{ ecs_task_definition.taskdefinition.revision }}"
# desired_count: 1
# state: present
# <<: *aws_connection_info
# register: ecs_service_creation_sr
# ignore_errors: yes
# - name: dump sr output
# debug: var=ecs_service_creation_sr
# - name: check that module returns success
# assert:
# that:
# - ecs_service_creation_sr.changed
- name: update ecs_service using REPLICA scheduling_strategy
ecs_service:
name: "{{ ecs_service_name }}-replica"
cluster: "{{ ecs_cluster_name }}"
load_balancers:
- targetGroupArn: "{{ elb_target_group_instance.target_group_arn }}"
containerName: "{{ ecs_task_name }}"
containerPort: "{{ ecs_task_container_port }}"
scheduling_strategy: "REPLICA"
task_definition: "{{ ecs_task_name }}:{{ ecs_task_definition.taskdefinition.revision }}"
desired_count: 1
state: present
<<: *aws_connection_info
register: ecs_service_creation_replica
- name: obtain facts for all ECS services in the cluster
ecs_service_facts:
cluster: "{{ ecs_cluster_name }}"
@ -728,6 +817,56 @@
ignore_errors: yes
register: ecs_service_scale_down
- name: scale down multifunction-test service
ecs_service:
name: "{{ ecs_service_name }}-mft"
cluster: "{{ ecs_cluster_name }}"
state: present
load_balancers:
- targetGroupArn: "{{ elb_target_group_instance.target_group_arn }}"
containerName: "{{ ecs_task_name }}"
containerPort: "{{ ecs_task_container_port }}"
task_definition: "{{ ecs_task_name }}:{{ ecs_task_definition.taskdefinition.revision }}"
desired_count: 0
<<: *aws_connection_info
ignore_errors: yes
register: ecs_service_scale_down
- name: scale down scheduling_strategy service
ecs_service:
name: "{{ ecs_service_name }}-replica"
cluster: "{{ ecs_cluster_name }}"
state: present
load_balancers:
- targetGroupArn: "{{ elb_target_group_instance.target_group_arn }}"
containerName: "{{ ecs_task_name }}"
containerPort: "{{ ecs_task_container_port }}"
task_definition: "{{ ecs_task_name }}:{{ ecs_task_definition.taskdefinition.revision }}"
desired_count: 0
<<: *aws_connection_info
ignore_errors: yes
register: ecs_service_scale_down
# until ansible supports service registries, the test for it can't run and this
# scale down is not needed
# - name: scale down service_registries service
# ecs_service:
# name: "{{ ecs_service_name }}-service-registries"
# cluster: "{{ ecs_cluster_name }}"
# state: present
# load_balancers:
# - targetGroupArn: "{{ elb_target_group_instance.target_group_arn }}"
# containerName: "{{ ecs_task_name }}"
# containerPort: "{{ ecs_task_container_port }}"
# task_definition: "{{ ecs_task_name }}:{{ ecs_task_definition.taskdefinition.revision }}"
# desired_count: 0
# <<: *aws_connection_info
# ignore_errors: yes
# register: ecs_service_scale_down
- name: scale down Fargate ECS service
ecs_service:
state: present
@ -761,6 +900,32 @@
<<: *aws_connection_info
ignore_errors: yes
- name: remove mft ecs service
ecs_service:
state: absent
cluster: "{{ ecs_cluster_name }}"
name: "{{ ecs_service_name }}-mft"
<<: *aws_connection_info
ignore_errors: yes
- name: remove scheduling_strategy ecs service
ecs_service:
state: absent
cluster: "{{ ecs_cluster_name }}"
name: "{{ ecs_service_name }}-replica"
<<: *aws_connection_info
ignore_errors: yes
# until ansible supports service registries, the test for it can't run and this
# removal is not needed
# - name: remove service_registries ecs service
# ecs_service:
# state: absent
# cluster: "{{ ecs_cluster_name }}"
# name: "{{ ecs_service_name }}-service-registries"
# <<: *aws_connection_info
# ignore_errors: yes
- name: remove fargate ECS service
ecs_service:
state: absent
@ -769,7 +934,7 @@
<<: *aws_connection_info
ignore_errors: yes
register: ecs_fargate_service_network_with_awsvpc
- name: remove ecs task definition
ecs_taskdefinition:
containers: "{{ ecs_task_containers }}"

View file

@ -12,6 +12,8 @@ trap 'rm -rf "${MYTMPDIR}"' EXIT
# but for the python3 tests we need virtualenv to use python3
PYTHON=${ANSIBLE_TEST_PYTHON_INTERPRETER:-python}
# Test graceful failure for older versions of botocore
virtualenv --system-site-packages --python "${PYTHON}" "${MYTMPDIR}/botocore-1.7.40"
source "${MYTMPDIR}/botocore-1.7.40/bin/activate"
@ -42,5 +44,5 @@ ansible-playbook -i ../../inventory -e @../../integration_config.yml -e @../../c
# Run full test suite
virtualenv --system-site-packages --python "${PYTHON}" "${MYTMPDIR}/botocore-recent"
source "${MYTMPDIR}/botocore-recent/bin/activate"
$PYTHON -m pip install 'botocore>=1.8.4' boto3
$PYTHON -m pip install 'botocore>=1.10.37' boto3 # version 1.10.37 for scheduling strategy
ansible-playbook -i ../../inventory -e @../../integration_config.yml -e @../../cloud-config-aws.yml -v playbooks/full_test.yml "$@"