Class: Hako::Schedulers::Ecs

Inherits:
Hako::Scheduler show all
Defined in:
lib/hako/schedulers/ecs.rb

Defined Under Namespace

Classes: NoTasksStarted

Constant Summary collapse

DEFAULT_CLUSTER = 
'default'
DEFAULT_FRONT_PORT = 
10000
MIN_WAIT_TASK_INTERVAL = 
1
MAX_WAIT_TASK_INTERVAL = 
120
TASK_ID_RE = 
/\(task ([\h-]+)\)\.\z/.freeze
RUN_TASK_INTERVAL = 
10
MIN_ASG_INTERVAL = 
1
MAX_ASG_INTERVAL = 
120

Instance Attribute Summary collapse

Instance Method Summary collapse

Methods inherited from Hako::Scheduler

#initialize, #validation_error!

Constructor Details

This class inherits a constructor from Hako::Scheduler

Instance Attribute Details

#task ⇒ Object (readonly)

Returns the value of attribute task.



29
30
31
 
# File 'lib/hako/schedulers/ecs.rb', line 29

def task
  @task
end

Instance Method Details

#call_rollback_started(task_definition, target_definition) ⇒ nil (private)

Parameters:

  • task_definition (Aws::ECS::Types::TaskDefinition)
  • target_definition (String)

Returns:

  • (nil) 


1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
 
# File 'lib/hako/schedulers/ecs.rb', line 1467

def call_rollback_started(task_definition, target_definition)
  current_app = task_definition.container_definitions.find { |c| c.name == 'app' }
  target_app = ecs_client.describe_task_definition(task_definition: target_definition).task_definition.container_definitions.find { |c| c.name == 'app' }
  if current_app && target_app
    @scripts.each { |script| script.rollback_started(current_app.image, target_app.image) }
  else
    Hako.logger.warn("Cannot find image_tag. current_app=#{current_app.inspect} target_app=#{target_app.inspect}. Skip calling Script#rollback_started")
  end
  nil
end

#check_secrets(container_definition) ⇒ nil (private)

Parameters:

  • container_definition (Hash)

Returns:

  • (nil) 


1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
 
# File 'lib/hako/schedulers/ecs.rb', line 1400

def check_secrets(container_definition)
  parameter_names = (container_definition[:secrets] || []).map { |secret| secret.fetch(:value_from) }
  invalid_parameter_names = parameter_names.each_slice(10).flat_map do |names|
    names = names.map do |name|
      if name.start_with?('arn:')
        name.slice(%r{:parameter(/.+)\z}, 1)
      else
        name
      end
    end
    ssm_client.get_parameters(names: names).invalid_parameters
  end
  unless invalid_parameter_names.empty?
    raise Error.new("Invalid parameters for secrets: #{invalid_parameter_names}")
  end

  nil
end

#configure(options) ⇒ Object

Parameters:

  • options (Hash<String, Object>) 


32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
 
# File 'lib/hako/schedulers/ecs.rb', line 32

def configure(options)
  @cluster = options.fetch('cluster', DEFAULT_CLUSTER)
  @desired_count = options.fetch('desired_count', nil)
  @region = options.fetch('region') { validation_error!('region must be set') }
  @role = options.fetch('role', nil)
  @task_role_arn = options.fetch('task_role_arn', nil)
  @ecs_elb_options = options.fetch('elb', nil)
  @ecs_elb_v2_options = options.fetch('elb_v2', nil)
  if @ecs_elb_options && @ecs_elb_v2_options
    validation_error!('Cannot specify both elb and elb_v2')
  end
  @network_mode = options.fetch('network_mode', nil)
  if @network_mode == 'awsvpc' && @ecs_elb_v2_options
    # awsvpc network mode requires ELB target group with target_type=ip
    @ecs_elb_v2_options['target_type'] = 'ip'
  end
  @dynamic_port_mapping = options.fetch('dynamic_port_mapping', @ecs_elb_options.nil?)
  @health_check_grace_period_seconds = options.fetch('health_check_grace_period_seconds') do
    @ecs_elb_options || @ecs_elb_v2_options ? 0 : nil
  end
  if options.key?('autoscaling')
    @autoscaling = EcsAutoscaling.new(options.fetch('autoscaling'), @region, ecs_elb_client, dry_run: @dry_run)
  end
  @autoscaling_group_for_oneshot = options.fetch('autoscaling_group_for_oneshot', nil)
  @autoscaling_topic_for_oneshot = options.fetch('autoscaling_topic_for_oneshot', nil)
  if @autoscaling_topic_for_oneshot && !@autoscaling_group_for_oneshot
    validation_error!('autoscaling_group_for_oneshot must be set when autoscaling_topic_for_oneshot is set')
  end
  @oneshot_notification_prefix = options.fetch('oneshot_notification_prefix', nil)
  if options.key?('deployment_configuration')
    @deployment_configuration = {}
    %i[maximum_percent minimum_healthy_percent].each do |key|
      @deployment_configuration[key] = options.fetch('deployment_configuration')[key.to_s]
    end
  else
    @deployment_configuration = nil
  end
  @placement_constraints = options.fetch('placement_constraints', [])
  @placement_strategy = options.fetch('placement_strategy', [])
  @scheduling_strategy = options.fetch('scheduling_strategy', nil)
  @execution_role_arn = options.fetch('execution_role_arn', nil)
  @cpu = options.fetch('cpu', nil)
  @memory = options.fetch('memory', nil)
  if options.key?('ephemeral_storage')
    ephemeral_storage = options.fetch('ephemeral_storage')
    if ephemeral_storage.key?('size_in_gi_b')
      @ephemeral_storage = {
        size_in_gi_b: ephemeral_storage.fetch('size_in_gi_b')
      }
    end
  end
  @requires_compatibilities = options.fetch('requires_compatibilities', nil)
  if options.key?('runtime_platform')
    runtime_platform = options.fetch('runtime_platform')
    @runtime_platform = {
      cpu_architecture: runtime_platform.fetch('cpu_architecture', nil),
      operating_system_family: runtime_platform.fetch('operating_system_family', nil),
    }
  end
  @launch_type = options.fetch('launch_type', nil)
  if options.key?('capacity_provider_strategy')
    @capacity_provider_strategy = options.fetch('capacity_provider_strategy').map do |strategy|
      {
        capacity_provider: strategy.fetch('capacity_provider'),
        weight: strategy.fetch('weight', nil),
        base: strategy.fetch('base', nil),
      }
    end
  end
  @platform_version = options.fetch('platform_version', nil)
  if options.key?('network_configuration')
    network_configuration = options.fetch('network_configuration')
    if network_configuration.key?('awsvpc_configuration')
      awsvpc_configuration = network_configuration.fetch('awsvpc_configuration')
      @network_configuration = {
        awsvpc_configuration: {
          subnets: awsvpc_configuration.fetch('subnets'),
          security_groups: awsvpc_configuration.fetch('security_groups', nil),
          assign_public_ip: awsvpc_configuration.fetch('assign_public_ip', nil),
        },
      }
    end
  end
  if options['service_discovery']
    @service_discovery = EcsServiceDiscovery.new(options.fetch('service_discovery'), @region, dry_run: @dry_run)
  end
  @tags = options.fetch('tags', {}).map { |k, v| { key: k, value: v.to_s } }

  @started_at = nil
  @container_instance_arn = nil
end

#create_definition(name, container) ⇒ Hash (private)

Parameters:

Returns:

  • (Hash) 


675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
 
# File 'lib/hako/schedulers/ecs.rb', line 675

def create_definition(name, container)
  environment = container.env.map { |k, v| { name: k, value: v } }
  {
    name: name,
    image: container.image_tag,
    cpu: container.cpu,
    memory: container.memory,
    memory_reservation: container.memory_reservation,
    links: container.links,
    port_mappings: container.port_mappings,
    essential: container.essential,
    environment: environment,
    secrets: container.secrets,
    docker_labels: container.docker_labels,
    mount_points: container.mount_points,
    entry_point: container.entry_point,
    command: container.command,
    privileged: container.privileged,
    linux_parameters: container.linux_parameters,
    depends_on: container.depends_on,
    volumes_from: container.volumes_from,
    user: container.user,
    log_configuration: container.log_configuration,
    health_check: container.health_check,
    ulimits: container.ulimits,
    extra_hosts: container.extra_hosts,
    readonly_root_filesystem: container.readonly_root_filesystem,
    docker_security_options: container.docker_security_options,
    system_controls: container.system_controls,
    repository_credentials: container.repository_credentials,
  }
end

#create_definitions(containers) ⇒ nil (private)

Parameters:

Returns:

  • (nil) 


593
594
595
596
597
 
# File 'lib/hako/schedulers/ecs.rb', line 593

def create_definitions(containers)
  containers.map do |name, container|
    create_definition(name, container)
  end
end

#create_initial_service(task_definition_arn, front_port) ⇒ Aws::ECS::Types::Service (private)

Parameters:

  • task_definition_arn (String)
  • front_port (Fixnum)

Returns:

  • (Aws::ECS::Types::Service) 


953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
 
# File 'lib/hako/schedulers/ecs.rb', line 953

def create_initial_service(task_definition_arn, front_port)
  params = {
    cluster: @cluster,
    service_name: @app_id,
    task_definition: task_definition_arn,
    role: @role,
    deployment_configuration: @deployment_configuration,
    placement_constraints: @placement_constraints,
    placement_strategy: @placement_strategy,
    scheduling_strategy: @scheduling_strategy,
    launch_type: @launch_type,
    capacity_provider_strategy: @capacity_provider_strategy,
    platform_version: @platform_version,
    network_configuration: @network_configuration,
    health_check_grace_period_seconds: @health_check_grace_period_seconds,
    propagate_tags: 'TASK_DEFINITION',
  }
  if @scheduling_strategy != 'DAEMON'
    params[:desired_count] = 0
  end
  if ecs_elb_client.find_or_create_load_balancer(front_port)
    params[:load_balancers] = [ecs_elb_client.load_balancer_params_for_service]
  end
  if @service_discovery
    @service_discovery.apply
    params[:service_registries] = @service_discovery.service_registries
  end
  ecs_client.create_service(**params).service
end

#deploy(containers) ⇒ nil

Parameters:

Returns:

  • (nil) 


126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
 
# File 'lib/hako/schedulers/ecs.rb', line 126

def deploy(containers)
  if @desired_count.nil? && @scheduling_strategy != 'DAEMON'
    validation_error!('desired_count must be set')
  end
  front_port = determine_front_port
  unless @dry_run
    ecs_elb_client.find_or_create_load_balancer(front_port)
  end
  @scripts.each { |script| script.deploy_started(containers, front_port) }
  definitions = create_definitions(containers)

  if @dry_run
    volumes_definition.each do |d|
      print_volume_definition_in_cli_format(d)
    end
    definitions.each do |d|
      print_definition_in_cli_format(d)
      check_secrets(d)
    end
    if @autoscaling
      @autoscaling.apply(Aws::ECS::Types::Service.new(cluster_arn: @cluster, service_name: @app_id))
    end
    ecs_elb_client.modify_attributes
    if @service_discovery
      @service_discovery.apply
    end
  else
    current_service = describe_service
    task_definition_changed, task_definition = register_task_definition(definitions)
    if task_definition_changed
      Hako.logger.info "Registered task definition: #{task_definition.task_definition_arn}"
    else
      Hako.logger.info "Task definition isn't changed: #{task_definition.task_definition_arn}"
    end
    current_service ||= create_initial_service(task_definition.task_definition_arn, front_port)
    service = update_service(current_service, task_definition.task_definition_arn)
    if service == :noop
      Hako.logger.info "Service isn't changed"
      if @autoscaling
        @autoscaling.apply(current_service)
      end
      ecs_elb_client.modify_attributes
      if @service_discovery
        @service_discovery.apply
      end
    else
      Hako.logger.info "Updated service: #{service.service_arn}"
      if @autoscaling
        @autoscaling.apply(service)
      end
      ecs_elb_client.modify_attributes
      if @service_discovery
        @service_discovery.apply
      end
      unless wait_for_ready(service)
        @scripts.each { |script| script.deploy_failed(containers, task_ids: @started_task_ids) }
        if task_definition_changed
          Hako.logger.error("Rolling back to #{current_service.task_definition}")
          update_service(service, current_service.task_definition)
          ecs_client.deregister_task_definition(task_definition: service.task_definition)
          Hako.logger.debug "Deregistered #{service.task_definition}"
        end
        raise Error.new('Deployment cancelled')
      end
    end
    Hako.logger.info 'Deployment completed'
  end
end

#describe_service ⇒ Aws::ECS::Types::Service? (private)

Returns:

  • (Aws::ECS::Types::Service, nil) 


419
420
421
422
423
424
 
# File 'lib/hako/schedulers/ecs.rb', line 419

def describe_service
  service = ecs_client.describe_services(cluster: @cluster, services: [@app_id]).services[0]
  if service && service.status != 'INACTIVE'
    service
  end
end

#describe_task_definition(family) ⇒ Object (private) 



665
666
667
668
669
670
 
# File 'lib/hako/schedulers/ecs.rb', line 665

def describe_task_definition(family)
  ecs_client.describe_task_definition(task_definition: family).task_definition
rescue Aws::ECS::Errors::ClientException
  # Task definition does not exist
  nil
end

#determine_front_port ⇒ Fixnum (private)

Returns:

  • (Fixnum) 


427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
 
# File 'lib/hako/schedulers/ecs.rb', line 427

def determine_front_port
  if @network_mode == 'awsvpc'
    # When networkMode=awsvpc, the host ports and container ports in port mappings must match
    return nil
  elsif @dynamic_port_mapping
    return 0
  end
  if @dry_run
    return DEFAULT_FRONT_PORT
  end

  service = describe_service
  if service
    find_front_port(service)
  else
    new_front_port
  end
end

#different_capacity_provider_strategy?(expected_strategy, actual_strategy) ⇒ Boolean (private)

Parameters:

  • expected_strategy (Hash, nil)
  • actual_strategy (Aws::ECS::Types::CapacityProviderStrategyItem, nil)

Returns:

  • (Boolean) 


1422
1423
1424
1425
1426
 
# File 'lib/hako/schedulers/ecs.rb', line 1422

def different_capacity_provider_strategy?(expected_strategy, actual_strategy)
  expected = (expected_strategy || []).map { |s| [s[:capacity_provider], s[:weight] || 0, s[:base] || 0] }.sort
  actual = (actual_strategy || []).map { |s| [s.capacity_provider, s.weight, s.base] }.sort
  expected != actual
end

#different_definition?(expected_container, actual_container) ⇒ Boolean (private)

Parameters:

  • expected_container (Hash)
  • actual_container (Aws::ECS::Types::ContainerDefinition)

Returns:

  • (Boolean) 


559
560
561
 
# File 'lib/hako/schedulers/ecs.rb', line 559

def different_definition?(expected_container, actual_container)
  EcsDefinitionComparator.new(expected_container).different?(actual_container)
end

#different_volume?(expected_volume, actual_volume) ⇒ Boolean (private)

Parameters:

  • expected_volume (Hash)
  • actual_volume (Aws::ECS::Types::Volume)

Returns:

  • (Boolean) 


552
553
554
 
# File 'lib/hako/schedulers/ecs.rb', line 552

def different_volume?(expected_volume, actual_volume)
  EcsVolumeComparator.new(expected_volume).different?(actual_volume)
end

#ec2_client ⇒ Aws::EC2::Client (private)

Returns:

  • (Aws::EC2::Client) 


399
400
401
 
# File 'lib/hako/schedulers/ecs.rb', line 399

def ec2_client
  @ec2_client ||= Aws::EC2::Client.new(region: @region)
end

#ecs_client ⇒ Aws::ECS::Client (private)

Returns:

  • (Aws::ECS::Client) 


394
395
396
 
# File 'lib/hako/schedulers/ecs.rb', line 394

def ecs_client
  @ecs_client ||= Aws::ECS::Client.new(region: @region)
end

#ecs_elb_client ⇒ EcsElb, EcsElbV2 (private)

Returns:



409
410
411
412
413
414
415
416
 
# File 'lib/hako/schedulers/ecs.rb', line 409

def ecs_elb_client
  @ecs_elb_client ||=
    if @ecs_elb_options
      EcsElb.new(@app_id, @region, @ecs_elb_options, dry_run: @dry_run)
    else
      EcsElbV2.new(@app_id, @region, @ecs_elb_v2_options, dry_run: @dry_run)
    end
end

#extract_task_id(message) ⇒ String? (private)

Parameters:

  • message (String)

Returns:

  • (String, nil) 


1062
1063
1064
 
# File 'lib/hako/schedulers/ecs.rb', line 1062

def extract_task_id(message)
  message.slice(TASK_ID_RE, 1)
end

#find_front_port(service) ⇒ Fixnum? (private)

Parameters:

  • service (Aws::ECS::Types::Service)

Returns:

  • (Fixnum, nil) 


470
471
472
473
474
475
476
477
478
479
 
# File 'lib/hako/schedulers/ecs.rb', line 470

def find_front_port(service)
  task_definition = ecs_client.describe_task_definition(task_definition: service.task_definition).task_definition
  container_definitions = {}
  task_definition.container_definitions.each do |c|
    container_definitions[c.name] = c
  end
  if container_definitions['front']
    container_definitions['front'].port_mappings[0].host_port
  end
end

#find_latest_event_id(events) ⇒ String? (private)

Parameters:

  • events (Array<Aws::ECS::Types::ServiceEvent>)

Returns:

  • (String, nil) 


1051
1052
1053
1054
1055
1056
1057
 
# File 'lib/hako/schedulers/ecs.rb', line 1051

def find_latest_event_id(events)
  if events.empty?
    nil
  else
    events[0].id
  end
end

#find_rollback_target(task_definition) ⇒ String (private)

Parameters:

  • (Aws::ECS::Types::TaskDefinition)

Returns:

  • (String) 


1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
 
# File 'lib/hako/schedulers/ecs.rb', line 1078

def find_rollback_target(task_definition)
  if task_definition.status != 'ACTIVE'
    raise 'Cannot find rollback target from INACTIVE task_definition!'
  end

  arn_found = false
  ecs_client.list_task_definitions(family_prefix: task_definition.family, status: 'ACTIVE', sort: 'DESC').each do |page|
    page.task_definition_arns.each do |arn|
      if arn_found
        return arn
      elsif arn == task_definition.task_definition_arn
        arn_found = true
      end
    end
  end

  raise "Unable to find rollback target. #{task_definition.task_definition_arn} is INACTIVE?"
end

#has_capacity?(task_definition, container_instances) ⇒ Boolean (private)

Parameters:

  • task_definition (Aws::ECS::Types::TaskDefinition)
  • container_instances (Array<Aws::ECS::Types::ContainerInstance>)

Returns:

  • (Boolean) 


1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
 
# File 'lib/hako/schedulers/ecs.rb', line 1198

def has_capacity?(task_definition, container_instances)
  required_cpu = task_definition.cpu && task_definition.cpu.to_i
  required_cpu ||= task_definition.container_definitions.inject(0) { |cpu, d| cpu + d.cpu }
  required_memory = task_definition.memory && task_definition.memory.to_i
  required_memory ||= task_definition.container_definitions.inject(0) { |memory, d| memory + (d.memory_reservation || d.memory) }
  container_instances.any? do |ci|
    cpu = ci.remaining_resources.find { |r| r.name == 'CPU' }.integer_value
    memory = ci.remaining_resources.find { |r| r.name == 'MEMORY' }.integer_value
    required_cpu <= cpu && required_memory <= memory
  end
end

#new_front_port ⇒ Fixnum (private)

Returns:

  • (Fixnum) 


447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
 
# File 'lib/hako/schedulers/ecs.rb', line 447

def new_front_port
  max_port = -1
  ecs_client.list_services(cluster: @cluster).each do |page|
    unless page.service_arns.empty?
      ecs_client.describe_services(cluster: @cluster, services: page.service_arns).services.each do |s|
        if s.status != 'INACTIVE'
          port = find_front_port(s)
          if port
            max_port = [max_port, port].max
          end
        end
      end
    end
  end
  if max_port == -1
    DEFAULT_FRONT_PORT
  else
    max_port + 1
  end
end

#on_no_tasks_started(task_definition) ⇒ Boolean (private)

Returns true if the capacity is reserved.

Parameters:

  • task_definition (Aws::ECS::Types::TaskDefinition)

Returns:

  • (Boolean) —

    true if the capacity is reserved



1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
 
# File 'lib/hako/schedulers/ecs.rb', line 1099

def on_no_tasks_started(task_definition)
  unless @autoscaling_group_for_oneshot
    return false
  end

  if @autoscaling_topic_for_oneshot
    try_scale_out_with_sns(task_definition)
  else
    try_scale_out_with_as(task_definition)
  end
end

#oneshot(containers, commands, env, no_wait: false, overrides: nil) ⇒ Integer

Returns exit code

Parameters:

  • containers (Hash<String, Container>)
  • commands (Array<String>)
  • env (Hash<String, String>)
  • no_wait (Boolean) (defaults to: false)
  • overrides (Hako::CLI::Oneshot::Overrides, nil) (defaults to: nil)

Returns:

  • (Integer) —

    Returns exit code



228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
 
# File 'lib/hako/schedulers/ecs.rb', line 228

def oneshot(containers, commands, env, no_wait: false, overrides: nil)
  definitions = create_definitions(containers)

  if @dry_run
    volumes_definition.each do |d|
      print_volume_definition_in_cli_format(d)
    end
    definitions.each do |d|
      if d[:name] == 'app'
        d[:command] = commands
      end
      print_definition_in_cli_format(d, additional_env: env, overrides: overrides)
      check_secrets(d)
    end
    0
  else
    updated, task_definition = register_task_definition_for_oneshot(definitions)
    if updated
      Hako.logger.info "Registered task definition: #{task_definition.task_definition_arn}"
    else
      Hako.logger.info "Task definition isn't changed: #{task_definition.task_definition_arn}"
    end
    @task = run_task(task_definition, commands, env, overrides)
    Hako.logger.info "Started task: #{@task.task_arn}"
    @scripts.each { |script| script.oneshot_started(self) }
    if no_wait
      info = { cluster: @cluster, task_arn: @task.task_arn }
      puts JSON.dump(info)
      0
    else
      wait_for_oneshot_finish
    end
  end
end

#overrides_option(commands, env, overrides) ⇒ Object (private)

Parameters:



753
754
755
756
757
758
759
760
761
762
763
764
765
766
 
# File 'lib/hako/schedulers/ecs.rb', line 753

def overrides_option(commands, env, overrides)
  {
    container_overrides: [
      {
        name: 'app',
        cpu: overrides&.app_cpu,
        memory: overrides&.app_memory,
        memory_reservation: overrides&.app_memory_reservation,
        command: commands,
        environment: env.map { |k, v| { name: k, value: v } },
      },
    ],
  }
end

#poll_task_status_from_ecs(task) ⇒ Hash<String, Aws::ECS::Types::Container> (private)

Parameters:

  • task (Aws::ECS::Types::Task)

Returns:

  • (Hash<String, Aws::ECS::Types::Container>) 


801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
 
# File 'lib/hako/schedulers/ecs.rb', line 801

def poll_task_status_from_ecs(task)
  task_arn = task.task_arn
  interval = 1
  loop do
    begin
      task = ecs_client.describe_tasks(cluster: @cluster, tasks: [task_arn]).tasks[0]
    rescue Aws::ECS::Errors::ThrottlingException => e
      Hako.logger.error(e)
      interval = [interval * 2, MAX_WAIT_TASK_INTERVAL].min
      Hako.logger.info("Retrying after #{interval} seconds...")
      sleep interval
      next
    end
    if task.nil?
      Hako.logger.debug "Task #{task_arn} could not be described"
      sleep 1
      next
    end

    if @container_instance_arn != task.container_instance_arn
      @container_instance_arn = task.container_instance_arn
      report_container_instance(@container_instance_arn)
    end
    unless @started_at
      @started_at = task.started_at
      if @started_at
        Hako.logger.info "Started at #{@started_at}"
      end
    end

    Hako.logger.debug "  status #{task.last_status}"

    if task.last_status == 'STOPPED'
      Hako.logger.info "Stopped at #{task.stopped_at} (reason: #{task.stopped_reason})"
      containers = {}
      task.containers.each do |c|
        containers[c.name] = c
      end
      return containers
    end
    interval = [interval / 2, MIN_WAIT_TASK_INTERVAL].max
    Hako.logger.debug("Waiting task with interval=#{interval}")
    sleep interval
  end
end

#poll_task_status_from_s3(task) ⇒ Hash<String, Aws::ECS::Types::Container> (private)

Get stopped container status from S3. The advantage is scalability; ecs:DescribeTasks is heavily rate-limited, but s3:GetObject is much more scalable. The JSON is supposed to be stored from Amazon ECS Event Stream. docs.aws.amazon.com/AmazonECS/latest/developerguide/cloudwatch_event_stream.html

Parameters:

  • task (Aws::ECS::Types::Task)

Returns:

  • (Hash<String, Aws::ECS::Types::Container>) 


854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
 
# File 'lib/hako/schedulers/ecs.rb', line 854

def poll_task_status_from_s3(task)
  s3 = Aws::S3::Client.new(region: @region)
  task_arn = task.task_arn
  uri = URI.parse(@oneshot_notification_prefix)
  prefix = uri.path.sub(%r{\A/}, '')
  started_key = "#{prefix}/#{task_arn}/started.json"
  stopped_key = "#{prefix}/#{task_arn}/stopped.json"

  loop do
    unless @started_at
      begin
        object = s3.get_object(bucket: uri.host, key: started_key)
      rescue Aws::S3::Errors::NoSuchKey
        Hako.logger.debug("  s3://#{uri.host}/#{started_key} doesn't exist")
      else
        json = JSON.parse(object.body.read)
        arn = json['detail']['containerInstanceArn']
        if @container_instance_arn != arn
          @container_instance_arn = arn
          report_container_instance(@container_instance_arn)
        end
        @started_at = Time.parse(json['detail']['startedAt'])
        if @started_at
          Hako.logger.info "Started at #{@started_at}"
        end
      end
    end

    begin
      object = s3.get_object(bucket: uri.host, key: stopped_key)
    rescue Aws::S3::Errors::NoSuchKey
      Hako.logger.debug("  s3://#{uri.host}/#{stopped_key} doesn't exist")
    else
      json = JSON.parse(object.body.read)
      task = Aws::Json::Parser.new(Aws::ECS::Client.api.operation('describe_tasks').output.shape.member(:tasks).shape.member).parse(json['detail'].to_json)
      Hako.logger.info "Stopped at #{task.stopped_at} (reason: #{task.stopped_reason})"
      containers = {}
      task.containers.each do |c|
        containers[c.name] = c
      end
      return containers
    end

    sleep 1
  end
end

#print_definition_in_cli_format(definition, additional_env: {}, overrides: nil) ⇒ nil (private)

Parameters:

Returns:

  • (nil) 


1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
 
# File 'lib/hako/schedulers/ecs.rb', line 1239

def print_definition_in_cli_format(definition, additional_env: {}, overrides: nil)
  cmd = %w[docker run]
  cmd << '--name' << definition.fetch(:name)

  if overrides && definition.fetch(:name) == 'app'
    cpu = overrides.app_cpu || definition.fetch(:cpu)
    cmd << '--cpu-shares' << cpu
    memory = overrides.app_memory || definition[:memory]
    if memory
      cmd << '--memory' << "#{memory}M"
    end
    memory_reservation = overrides.app_memory_reservation || definition[:memory_reservation]
    if memory_reservation
      cmd << '--memory-reservation' << "#{memory_reservation}M"
    end
  else
    cmd << '--cpu-shares' << definition.fetch(:cpu)
    if definition[:memory]
      cmd << '--memory' << "#{definition[:memory]}M"
    end
    if definition[:memory_reservation]
      cmd << '--memory-reservation' << "#{definition[:memory_reservation]}M"
    end
  end

  definition.fetch(:links).each do |link|
    cmd << '--link' << link
  end
  definition.fetch(:port_mappings).each do |port_mapping|
    cmd << '--publish' << "#{port_mapping.fetch(:host_port)}:#{port_mapping.fetch(:container_port)}"
  end
  definition.fetch(:docker_labels).each do |key, val|
    if key != 'cc.wanko.hako.version'
      cmd << '--label' << "#{key}=#{val}"
    end
  end
  definition.fetch(:mount_points).each do |mount_point|
    source_volume = mount_point.fetch(:source_volume)
    v = volumes_definition.find { |d| d[:name] == source_volume }
    raise "Could not find volume #{source_volume}" unless v

    source = v.dig(:host, :source_path) || source_volume
    cmd << '--volume' << "#{source}:#{mount_point.fetch(:container_path)}#{mount_point[:read_only] ? ':ro' : ''}"
  end
  definition.fetch(:volumes_from).each do |volumes_from|
    cmd << '--volumes-from' << "#{volumes_from.fetch(:source_container)}#{volumes_from[:read_only] ? ':ro' : ''}"
  end
  if definition[:privileged]
    cmd << '--privileged'
  end
  if definition[:linux_parameters]
    if definition[:linux_parameters][:capabilities]
      cp = definition[:linux_parameters][:capabilities]
      %i[add drop].each do |a_or_d|
        cp[a_or_d]&.each do |c|
          cmd << "--cap-#{a_or_d}=#{c}"
        end
      end
    end

    if definition[:linux_parameters][:devices]
      devs = definition[:linux_parameters][:devices]
      devs.each do |dev|
        opts = dev[:host_path]
        opts += ":#{dev[:container_path]}" if dev[:container_path]
        if dev[:permissions]
          dev[:permissions].each do |permission|
            opts += permission[0] if %w[read write mknod].include?(permission)
          end
        end
        cmd << "--device=#{opts}"
      end
    end

    if definition[:linux_parameters][:init_process_enabled]
      cmd << '--init'
    end

    if definition[:linux_parameters][:shared_memory_size]
      cmd << '--shm-size' << "#{definition[:linux_parameters][:shared_memory_size]}m"
    end

    definition[:linux_parameters].fetch(:tmpfs, []).each do |tmpfs|
      options = ["size=#{tmpfs[:size]}m"].concat(tmpfs[:mount_options])
      cmd << '--tmpfs' << "#{tmpfs[:container_path]}:#{options.join(',')}"
    end
  end
  if definition[:user]
    cmd << '--user' << definition[:user]
  end
  if definition[:readonly_root_filesystem]
    cmd << '--read-only'
  end
  (definition[:docker_security_options] || []).each do |docker_security_option|
    cmd << '--security-opt' << docker_security_option
  end
  if definition[:entry_point]
    cmd << '--entrypoint' << definition[:entry_point]
  end
  if definition[:system_controls]
    definition.fetch(:system_controls).each do |system_control|
      cmd << '--sysctl' << "#{system_control.fetch(:namespace)}=#{system_control.fetch(:value)}"
    end
  end
  if definition[:health_check]
    if definition[:health_check][:command]
      health_check_command_type = definition[:health_check][:command][0]
      case health_check_command_type
      when 'NONE'
        cmd << '--no-healthcheck'
      when 'CMD', 'CMD-SHELL'
        health_check_command = definition[:health_check][:command][1..].join(' ')
        cmd << '--health-cmd' << health_check_command.inspect
      else
        raise "Health check command type #{health_check_command_type} is not supported. CMD, CMD-SHELL and NONE are supported."
      end
    end
    if definition[:health_check][:retries]
      cmd << '--health-retries' << definition[:health_check][:retries]
    end
    %i[interval timeout start_period].each do |property|
      if definition[:health_check][property]
        cmd << "--health-#{property}" << "#{definition[:health_check][property]}s"
      end
    end
  end

  cmd << "\\\n  "
  definition.fetch(:environment).each do |env|
    name = env.fetch(:name)
    value = env.fetch(:value)
    # additional_env (given in command line) has priority over env (declared in definition file)
    unless additional_env.key?(name)
      cmd << '--env' << "#{name}=#{value}"
      cmd << "\\\n  "
    end
  end
  (definition[:secrets] || []).each do |secret|
    name = secret.fetch(:name)
    # additional_env (given in command line) has priority over secrets (declared in definition file)
    unless additional_env.key?(name)
      cmd << '--env' << "#{name}=secret:#{secret.fetch(:value_from)}"
      cmd << "\\\n  "
    end
  end
  additional_env.each do |name, value|
    cmd << '--env' << "#{name}=#{value}"
    cmd << "\\\n  "
  end

  cmd << definition.fetch(:image)
  if definition[:command]
    cmd << "\\\n  "
    cmd += definition[:command]
  end
  puts cmd.join(' ')
  nil
end

#print_volume_definition_in_cli_format(definition) ⇒ nil (private)

Parameters:

  • definition (Hash)

Returns:

  • (nil) 


1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
 
# File 'lib/hako/schedulers/ecs.rb', line 1212

def print_volume_definition_in_cli_format(definition)
  return if definition.dig(:docker_volume_configuration, :autoprovision)
  # From version 1.20.0 of ECS agent, a local volume is provisioned when
  # 'host' is specified without 'source_path'.
  return if definition.dig(:host, :source_path)

  cmd = %w[docker volume create]
  if (configuration = definition[:docker_volume_configuration])
    if configuration[:driver]
      cmd << '--driver' << configuration[:driver]
    end
    (configuration[:driver_opts] || {}).each do |k, v|
      cmd << '--opt' << "#{k}=#{v}"
    end
    (configuration[:labels] || {}).each do |k, v|
      cmd << '--label' << "#{k}=#{v}"
    end
  end
  cmd << definition[:name]
  puts cmd.join(' ')
  nil
end

#register_task_definition(definitions) ⇒ Array<Boolean, Aws::ECS::Types::TaskDefinition> (private)

Parameters:

  • definitions (Array<Hash>)

Returns:

  • (Array<Boolean, Aws::ECS::Types::TaskDefinition>) 


565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
 
# File 'lib/hako/schedulers/ecs.rb', line 565

def register_task_definition(definitions)
  current_task_definition = describe_task_definition(@app_id)
  if current_task_definition
    current_tags = ecs_client.list_tags_for_resource(resource_arn: current_task_definition.task_definition_arn).tags
  end
  if task_definition_changed?(definitions, current_task_definition, current_tags)
    new_task_definition = ecs_client.register_task_definition(
      family: @app_id,
      task_role_arn: @task_role_arn,
      execution_role_arn: @execution_role_arn,
      network_mode: @network_mode,
      container_definitions: definitions,
      volumes: volumes_definition,
      requires_compatibilities: @requires_compatibilities,
      runtime_platform: @runtime_platform,
      cpu: @cpu,
      memory: @memory,
      ephemeral_storage: @ephemeral_storage,
      tags: @tags.empty? ? nil : @tags,
    ).task_definition
    [true, new_task_definition]
  else
    [false, current_task_definition]
  end
end

#register_task_definition_for_oneshot(definitions) ⇒ Array<Boolean, Aws::ECS::Types::TaskDefinition] (private)

Returns Array<Boolean, Aws::ECS::Types::TaskDefinition].

Parameters:

  • definitions (Array<Hash>)

Returns:

  • (Array<Boolean, Aws::ECS::Types::TaskDefinition]) —

    Array<Boolean, Aws::ECS::Types::TaskDefinition]

Raises:



601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
 
# File 'lib/hako/schedulers/ecs.rb', line 601

def register_task_definition_for_oneshot(definitions)
  10.times do |i|
    family = "#{@app_id}-oneshot"
    current_task_definition = describe_task_definition(family)
    if current_task_definition
      current_tags = ecs_client.list_tags_for_resource(resource_arn: current_task_definition.task_definition_arn).tags
    end
    if task_definition_changed?(definitions, current_task_definition, current_tags)
      new_task_definition = ecs_client.register_task_definition(
        family: family,
        task_role_arn: @task_role_arn,
        execution_role_arn: @execution_role_arn,
        network_mode: @network_mode,
        container_definitions: definitions,
        volumes: volumes_definition,
        requires_compatibilities: @requires_compatibilities,
        runtime_platform: @runtime_platform,
        cpu: @cpu,
        memory: @memory,
        ephemeral_storage: @ephemeral_storage,
        tags: @tags.empty? ? nil : @tags,
      ).task_definition
      return [true, new_task_definition]
    else
      return [false, current_task_definition]
    end
  rescue Aws::ECS::Errors::ClientException => e
    if e.message.include?('Too many concurrent attempts to create a new revision of the specified family')
      Hako.logger.error(e.message)
      interval = 2**i + rand(0.0..10.0)
      Hako.logger.error("Retrying register_task_definition_for_oneshot after #{interval} seconds")
      sleep(interval)
    else
      raise e
    end
  end
  raise Error.new('Unable to register task definition for oneshot due to too many client errors')
end

#remove ⇒ nil

Returns:

  • (nil) 


351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
 
# File 'lib/hako/schedulers/ecs.rb', line 351

def remove
  service = describe_service
  if service
    if @dry_run
      unless service.scheduling_strategy == 'DAEMON'
        Hako.logger.info "ecs_client.update_service(cluster: #{service.cluster_arn}, service: #{service.service_arn}, desired_count: 0)"
      end
      Hako.logger.info "ecs_client.delete_service(cluster: #{service.cluster_arn}, service: #{service.service_arn})"
    else
      unless service.scheduling_strategy == 'DAEMON'
        ecs_client.update_service(cluster: service.cluster_arn, service: service.service_arn, desired_count: 0)
      end
      ecs_client.delete_service(cluster: service.cluster_arn, service: service.service_arn)
      Hako.logger.info "#{service.service_arn} is deleted"
    end
    unless service.service_registries.empty?
      @service_discovery.remove(service.service_registries)
    end
  else
    puts "Service #{@app_id} doesn't exist"
  end

  ecs_elb_client.destroy
end

#report_container_instance(container_instance_arn) ⇒ nil (private)

Parameters:

  • container_instance_arn (String)

Returns:

  • (nil) 


903
904
905
906
 
# File 'lib/hako/schedulers/ecs.rb', line 903

def report_container_instance(container_instance_arn)
  container_instance = ecs_client.describe_container_instances(cluster: @cluster, container_instances: [container_instance_arn]).container_instances[0]
  Hako.logger.info "Container instance is #{container_instance_arn} (#{container_instance.ec2_instance_id})"
end

#report_task_diagnostics(task) ⇒ nil (private)

Parameters:

  • task (Aws::ECS::Types::Task)

Returns:

  • (nil) 


1068
1069
1070
1071
1072
1073
1074
 
# File 'lib/hako/schedulers/ecs.rb', line 1068

def report_task_diagnostics(task)
  Hako.logger.error("task_definition_arn=#{task.task_definition_arn} last_status=#{task.last_status}")
  Hako.logger.error("  stopped_reason: #{task.stopped_reason}")
  task.containers.sort_by(&:name).each do |container|
    Hako.logger.error("    Container #{container.name}: last_status=#{container.last_status} exit_code=#{container.exit_code.inspect} reason=#{container.reason.inspect}")
  end
end

#rollback ⇒ Object 



195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
 
# File 'lib/hako/schedulers/ecs.rb', line 195

def rollback
  current_service = describe_service
  unless current_service
    Hako.logger.error 'Unable to find service'
    exit 1
  end

  task_definition = ecs_client.describe_task_definition(task_definition: current_service.task_definition).task_definition
  current_definition = "#{task_definition.family}:#{task_definition.revision}"
  target_definition = find_rollback_target(task_definition)
  Hako.logger.info "Current task definition is #{current_definition}. Rolling back to #{target_definition}"
  call_rollback_started(task_definition, target_definition)

  if @dry_run
    Hako.logger.info 'Deployment completed (dry-run)'
  else
    service = ecs_client.update_service(cluster: current_service.cluster_arn, service: current_service.service_arn, task_definition: target_definition).service
    Hako.logger.info "Updated service: #{service.service_arn}"

    deregistered_definition = ecs_client.deregister_task_definition(task_definition: current_definition).task_definition
    Hako.logger.debug "Deregistered #{deregistered_definition.task_definition_arn}"

    wait_for_ready(service)
    Hako.logger.info 'Deployment completed'
  end
end

#run_task(task_definition, commands, env, overrides) ⇒ Aws::ECS::Types::Task (private)

Parameters:

  • task_definition (Aws::ECS::Types::TaskDefinition)
  • commands (Array<String>)
  • env (Hash<String, String>)
  • overrides (Hako::CLI::Oneshot::Overrides)

Returns:

  • (Aws::ECS::Types::Task) 


713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
 
# File 'lib/hako/schedulers/ecs.rb', line 713

def run_task(task_definition, commands, env, overrides)
  result = ecs_client.run_task(
    cluster: @cluster,
    task_definition: task_definition.task_definition_arn,
    overrides: overrides_option(commands, env, overrides),
    count: 1,
    placement_constraints: @placement_constraints,
    started_by: 'hako oneshot',
    launch_type: @launch_type,
    capacity_provider_strategy: @capacity_provider_strategy,
    platform_version: @platform_version,
    network_configuration: @network_configuration,
    propagate_tags: 'TASK_DEFINITION',
  )
  result.failures.each do |failure|
    Hako.logger.error("#{failure.arn} #{failure.reason}")
  end
  if result.tasks.empty?
    raise NoTasksStarted.new('No tasks started')
  end

  result.tasks[0]
rescue Aws::ECS::Errors::InvalidParameterException => e
  if e.message == 'No Container Instances were found in your cluster.' && on_no_tasks_started(task_definition)
    retry
  else
    raise e
  end
rescue NoTasksStarted => e
  if on_no_tasks_started(task_definition)
    retry
  else
    raise e
  end
end

#service_changed?(service, params) ⇒ Boolean (private)

Parameters:

  • service (Aws::ECS::Types::Service)
  • params (Hash)

Returns:

  • (Boolean) 


986
987
988
 
# File 'lib/hako/schedulers/ecs.rb', line 986

def service_changed?(service, params)
  EcsServiceComparator.new(params).different?(service)
end

#ssm_client ⇒ Aws::SSM::Client (private)

Returns:

  • (Aws::SSM::Client) 


404
405
406
 
# File 'lib/hako/schedulers/ecs.rb', line 404

def ssm_client
  @ssm_client ||= Aws::SSM::Client.new(region: @region)
end

#status ⇒ nil

Returns:

  • (nil) 


273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
 
# File 'lib/hako/schedulers/ecs.rb', line 273

def status
  service = describe_service
  unless service
    puts 'Unavailable'
    exit 1
  end

  unless service.load_balancers.empty?
    puts 'Load balancer:'
    ecs_elb_client.show_status(service.load_balancers[0])
  end

  puts 'Deployments:'
  service.deployments.each do |d|
    abbrev_task_definition = d.task_definition.slice(%r{task-definition/(.+)\z}, 1)
    puts "  [#{d.status}] #{abbrev_task_definition} desired_count=#{d.desired_count}, pending_count=#{d.pending_count}, running_count=#{d.running_count}"
  end

  puts 'Tasks:'
  ecs_client.list_tasks(cluster: @cluster, service_name: service.service_arn).each do |page|
    unless page.task_arns.empty?
      tasks = ecs_client.describe_tasks(cluster: @cluster, tasks: page.task_arns).tasks
      container_instance_arns = tasks.map(&:container_instance_arn).compact
      container_instances = {}
      unless container_instance_arns.empty?
        ecs_client.describe_container_instances(cluster: @cluster, container_instances: container_instance_arns).container_instances.each do |ci|
          container_instances[ci.container_instance_arn] = ci
        end
      end
      ec2_instances = {}
      unless container_instances.empty?
        ec2_client.describe_instances(instance_ids: container_instances.values.map(&:ec2_instance_id)).reservations.each do |r|
          r.instances.each do |i|
            ec2_instances[i.instance_id] = i
          end
        end
      end
      tasks.each do |task|
        task_id = task.task_arn.slice(%r{task/(.+)\z}, 1)
        task_definition = task.task_definition_arn.slice(%r{task-definition/(.+)\z}, 1)
        print "  [#{task.last_status}] #{task_id} task_definition=#{task_definition}, desired_status=#{task.desired_status}"
        ci = container_instances[task.container_instance_arn]
        if ci
          print ", instance_id=#{ci.ec2_instance_id}"
          instance = ec2_instances[ci.ec2_instance_id]
          if instance
            name_tag = instance.tags.find { |t| t.key == 'Name' }
            if name_tag
              print " (#{name_tag.value})"
            end
          end
        end
        puts
      end
    end
  end

  puts 'Events:'
  service.events.first(10).each do |e|
    puts "  #{e.created_at}: #{e.message}"
  end

  if @autoscaling
    puts 'Autoscaling:'
    @autoscaling.status(service)
  else
    puts 'Autoscaling: No'
  end

  if service.service_registries.empty?
    puts 'Service Discovery: No'
  else
    puts 'Service Discovery:'
    @service_discovery.status(service.service_registries)
  end
end

#stop ⇒ nil

Returns:

  • (nil) 


377
378
379
380
381
382
383
384
385
386
387
388
389
 
# File 'lib/hako/schedulers/ecs.rb', line 377

def stop
  service = describe_service
  if service
    if @dry_run
      Hako.logger.info("ecs_client.update_service(cluster: #{service.cluster_arn}, service: #{service.service_arn}, desired_count: 0)")
    else
      ecs_client.update_service(cluster: service.cluster_arn, service: service.service_arn, desired_count: 0)
      Hako.logger.info("#{service.service_arn} is stopped")
    end
  else
    puts "Service #{@app_id} doesn't exist"
  end
end

#stop_oneshot ⇒ nil

Returns:

  • (nil) 


264
265
266
267
268
269
270
 
# File 'lib/hako/schedulers/ecs.rb', line 264

def stop_oneshot
  if @task
    Hako.logger.warn "Stopping #{@task.task_arn}"
    ecs_client.stop_task(cluster: @cluster, task: @task.task_arn, reason: 'Stopped by hako stop_oneshot')
    wait_for_oneshot_finish
  end
end

#task_definition_changed?(desired_definitions, actual_definition, actual_tags) ⇒ Array<Boolean> (private)

Parameters:

  • desired_definitions (Array<Hash>)
  • actual_definition (Aws::ECS::Types::TaskDefinition)
  • actual_tags (Array<Aws::ECS::Types::Tag>)

Returns:

  • (Array<Boolean>) 


485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
 
# File 'lib/hako/schedulers/ecs.rb', line 485

def task_definition_changed?(desired_definitions, actual_definition, actual_tags)
  if @force
    return true
  end
  unless actual_definition
    # Initial deployment
    return true
  end

  actual_volume_definitions = {}
  actual_definition.volumes.each do |v|
    actual_volume_definitions[v.name] = v
  end
  container_definitions = {}
  actual_definition.container_definitions.each do |c|
    container_definitions[c.name] = c
  end

  if actual_definition.task_role_arn != @task_role_arn
    return true
  end
  if volumes_definition.any? { |definition| different_volume?(definition, actual_volume_definitions.delete(definition[:name])) }
    return true
  end
  unless actual_volume_definitions.empty?
    return true
  end
  if desired_definitions.any? { |definition| different_definition?(definition, container_definitions.delete(definition[:name])) }
    return true
  end
  unless container_definitions.empty?
    return true
  end
  if actual_definition.cpu != @cpu
    return true
  end
  if actual_definition.memory != @memory
    return true
  end
  if actual_definition.network_mode != @network_mode
    return true
  end
  if actual_definition.execution_role_arn != @execution_role_arn
    return true
  end
  if actual_definition.requires_compatibilities != @requires_compatibilities
    return true
  end
  if actual_definition.runtime_platform != @runtime_platform
    return true
  end
  if actual_definition.ephemeral_storage != @ephemeral_storage
    return true
  end

  actual_tags_set = Set.new(actual_tags.map { |t| { key: t.key, value: t.value } })
  tags_set = Set.new(@tags)
  if actual_tags_set != tags_set
    return true
  end

  false
end

#try_scale_out_with_as(task_definition) ⇒ Object (private) 



1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
 
# File 'lib/hako/schedulers/ecs.rb', line 1135

def try_scale_out_with_as(task_definition)
  autoscaling = Aws::AutoScaling::Client.new(region: @region)
  interval = MIN_ASG_INTERVAL
  Hako.logger.info("Unable to start tasks. Start trying scaling out '#{@autoscaling_group_for_oneshot}'")
  loop do
    begin
      asg = autoscaling.describe_auto_scaling_groups(auto_scaling_group_names: [@autoscaling_group_for_oneshot]).auto_scaling_groups[0]
    rescue Aws::AutoScaling::Errors::Throttling => e
      Hako.logger.error(e)
      interval = [interval * 2, MAX_ASG_INTERVAL].min
      Hako.logger.info("Retrying after #{interval} seconds...")
      sleep interval
      next
    end
    unless asg
      raise Error.new("AutoScaling Group '#{@autoscaling_group_for_oneshot}' does not exist")
    end

    container_instances = ecs_client.list_container_instances(cluster: @cluster).flat_map do |c|
      if c.container_instance_arns.empty?
        []
      else
        ecs_client.describe_container_instances(cluster: @cluster, container_instances: c.container_instance_arns).container_instances.select do |container_instance|
          container_instance.agent_connected && container_instance.status == 'ACTIVE'
        end
      end
    end
    if has_capacity?(task_definition, container_instances)
      Hako.logger.info("There's remaining capacity. Start retrying...")
      return true
    end

    interval = [interval / 2, MIN_ASG_INTERVAL].max
    # Check autoscaling group health
    current = asg.instances.count { |i| i.lifecycle_state == 'InService' }
    if asg.desired_capacity != current
      Hako.logger.debug("#{asg.auto_scaling_group_name} isn't in desired state. desired_capacity=#{asg.desired_capacity} in-service instances=#{current}. Retry after #{interval} seconds")
      sleep interval
      next
    end

    # Check out-of-service instances
    out_instances = asg.instances.map(&:instance_id)
    container_instances.each do |ci|
      out_instances.delete(ci.ec2_instance_id)
    end
    unless out_instances.empty?
      Hako.logger.debug("There's instances that is running but not registered as container instances: #{out_instances}. Retry after #{interval} seconds")
      sleep interval
      next
    end

    # Scale out
    desired = current + 1
    Hako.logger.info("Increment desired_capacity of #{asg.auto_scaling_group_name} from #{current} to #{desired}")
    autoscaling.set_desired_capacity(auto_scaling_group_name: asg.auto_scaling_group_name, desired_capacity: desired)
    sleep interval
  end
end

#try_scale_out_with_sns(task_definition) ⇒ Object (private) 



1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
 
# File 'lib/hako/schedulers/ecs.rb', line 1112

def try_scale_out_with_sns(task_definition)
  required_cpu = task_definition.cpu && task_definition.cpu.to_i
  required_cpu ||= task_definition.container_definitions.inject(0) { |cpu, d| cpu + d.cpu }
  required_memory = task_definition.memory && task_definition.memory.to_i
  required_memory ||= task_definition.container_definitions.inject(0) { |memory, d| memory + (d.memory_reservation || d.memory) }
  @hako_task_id ||= SecureRandom.uuid
  message = JSON.dump(
    group_name: @autoscaling_group_for_oneshot,
    cluster: @cluster,
    cpu: required_cpu,
    memory: required_memory,
    hako_task_id: @hako_task_id,
  )
  Hako.logger.info("Unable to start tasks. Publish message to #{@autoscaling_topic_for_oneshot}: #{message}")
  sns_client = Aws::SNS::Client.new(region: @region)
  resp = sns_client.publish(topic_arn: @autoscaling_topic_for_oneshot, message: message)
  Hako.logger.info("Sent message_id=#{resp.message_id}")
  sleep(RUN_TASK_INTERVAL)
  true
end

#update_service(current_service, task_definition_arn) ⇒ Aws::ECS::Types::Service, Symbol (private)

Parameters:

  • task_definition_arn (Aws::ECS::Types::Service)
  • task_definition_arn (String)

Returns:

  • (Aws::ECS::Types::Service, Symbol) 


911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
 
# File 'lib/hako/schedulers/ecs.rb', line 911

def update_service(current_service, task_definition_arn)
  params = {
    cluster: @cluster,
    service: @app_id,
    desired_count: @desired_count,
    task_definition: task_definition_arn,
    deployment_configuration: @deployment_configuration,
    capacity_provider_strategy: @capacity_provider_strategy,
    platform_version: @platform_version,
    network_configuration: @network_configuration,
    health_check_grace_period_seconds: @health_check_grace_period_seconds,
  }
  if @autoscaling
    # Keep current desired_count if autoscaling is enabled
    params[:desired_count] = current_service.desired_count
  end
  # Copy the current capacity provider strategy in order to avoid a
  # perpetual diff when the service is created with no strategy to use the
  # cluster's default capacity provider strategy, which results in the
  # strategy being set to the default strategy at that moment.
  # It is not allowed to update the service to use the cluster's default
  # capacity provider strategy when it is using a non-default capacity
  # provider strategy.
  params[:capacity_provider_strategy] ||= current_service.capacity_provider_strategy&.map(&:to_h)
  if different_capacity_provider_strategy?(params[:capacity_provider_strategy], current_service.capacity_provider_strategy)
    # Switching from launch type to capacity provider strategy or making
    # a change to a capacity provider strategy requires to force a new
    # deployment.
    params[:force_new_deployment] = true
  end
  warn_placement_policy_change(current_service)
  warn_service_registries_change(current_service)
  if service_changed?(current_service, params)
    ecs_client.update_service(**params).service
  else
    :noop
  end
end

#volumes_definition ⇒ Array<Hash> (private)

Returns:

  • (Array<Hash>) 


641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
 
# File 'lib/hako/schedulers/ecs.rb', line 641

def volumes_definition
  @volumes_definition ||= @volumes.map do |name, volume|
    definition = { name: name }
    if volume.key?('docker_volume_configuration')
      configuration = volume['docker_volume_configuration']
      definition[:docker_volume_configuration] = {
        autoprovision: configuration['autoprovision'],
        driver: configuration['driver'],
        # ECS API doesn't allow 'driver_opts' to be an empty hash.
        driver_opts: configuration['driver_opts'],
        # ECS API doesn't allow 'labels' to be an empty hash.
        labels: configuration['labels'],
        scope: configuration['scope'],
      }
    else
      # When neither 'host' nor 'docker_volume_configuration' is
      # specified, ECS API treats it as if 'host' is specified without
      # 'source_path'.
      definition[:host] = { source_path: volume['source_path'] }
    end
    definition
  end
end

#wait_for_oneshot_finish ⇒ Fixnum (private)

Returns:

  • (Fixnum) 


769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
 
# File 'lib/hako/schedulers/ecs.rb', line 769

def wait_for_oneshot_finish
  containers = wait_for_task(@task)
  @task = nil
  Hako.logger.info 'Oneshot task finished'
  exit_code = 127
  containers.each do |name, container|
    if container.exit_code.nil?
      Hako.logger.info "#{name} has stopped without exit_code: reason=#{container.reason}"
    else
      Hako.logger.info "#{name} has stopped with exit_code=#{container.exit_code}"
      if name == 'app'
        exit_code = container.exit_code
      end
    end
  end
  exit_code
end

#wait_for_ready(service) ⇒ Boolean (private)

Parameters:

  • service (Aws::ECS::Types::Service)

Returns:

  • (Boolean) 


992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
 
# File 'lib/hako/schedulers/ecs.rb', line 992

def wait_for_ready(service)
  latest_event_id = find_latest_event_id(service.events)
  Hako.logger.debug "  latest_event_id=#{latest_event_id}"
  started_at =
    if @timeout
      Process.clock_gettime(Process::CLOCK_MONOTONIC)
    end

  @started_task_ids = []

  loop do
    if started_at
      if Process.clock_gettime(Process::CLOCK_MONOTONIC) - started_at > @timeout
        Hako.logger.error('Timed out')
        return false
      end
    end

    s = ecs_client.describe_services(cluster: service.cluster_arn, services: [service.service_arn]).services[0]
    if s.nil?
      Hako.logger.debug "Service #{service.service_arn} could not be described"
      sleep 1
      next
    end
    s.events.each do |e|
      if e.id == latest_event_id
        break
      end

      Hako.logger.info "#{e.created_at}: #{e.message}"
      task_id = extract_task_id(e.message)
      if task_id && e.message.include?(' has started ')
        @started_task_ids << task_id
      end
    end
    latest_event_id = find_latest_event_id(s.events)
    Hako.logger.debug "  latest_event_id=#{latest_event_id}, deployments=#{s.deployments}"
    no_active = s.deployments.all? { |d| d.status != 'ACTIVE' }
    primary = s.deployments.find { |d| d.status == 'PRIMARY' }
    if primary.desired_count * 2 < @started_task_ids.size
      Hako.logger.error('Some started tasks are stopped. It seems new deployment is failing to start')
      @started_task_ids.each_slice(100) do |task_ids|
        ecs_client.describe_tasks(cluster: service.cluster_arn, tasks: task_ids).tasks.each do |task|
          report_task_diagnostics(task)
        end
      end
      return false
    end
    primary_ready = primary && primary.running_count == primary.desired_count
    if no_active && primary_ready
      return true
    else
      sleep 1
    end
  end
end

#wait_for_task(task) ⇒ Hash<String, Aws::ECS::Types::Container> (private)

Parameters:

  • task (Aws::ECS::Types::Task)

Returns:

  • (Hash<String, Aws::ECS::Types::Container>) 


789
790
791
792
793
794
795
 
# File 'lib/hako/schedulers/ecs.rb', line 789

def wait_for_task(task)
  if @oneshot_notification_prefix
    poll_task_status_from_s3(task)
  else
    poll_task_status_from_ecs(task)
  end
end

#warn_placement_policy_change(service) ⇒ nil (private)

Parameters:

  • service (Aws::ECS::Types::Service)

Returns:

  • (nil) 


1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
 
# File 'lib/hako/schedulers/ecs.rb', line 1430

def warn_placement_policy_change(service)
  placement_constraints = service.placement_constraints.map do |c|
    h = { 'type' => c.type }
    unless c.expression.nil?
      h['expression'] = c.expression
    end
    h
  end
  if @placement_constraints != placement_constraints
    Hako.logger.warn "Ignoring updated placement_constraints in the configuration, because AWS doesn't allow updating them for now."
  end

  placement_strategy = service.placement_strategy.map do |s|
    h = { 'type' => s.type }
    unless s.field.nil?
      h['field'] = s.field.downcase
    end
    h
  end
  if @placement_strategy != placement_strategy
    Hako.logger.warn "Ignoring updated placement_strategy in the configuration, because AWS doesn't allow updating them for now."
  end
end

#warn_service_registries_change(service) ⇒ void (private)

This method returns an undefined value.

Parameters:

  • service (Aws::ECS::Types::Service) 


1456
1457
1458
1459
1460
1461
1462
 
# File 'lib/hako/schedulers/ecs.rb', line 1456

def warn_service_registries_change(service)
  actual_service_registries = service.service_registries.sort_by(&:registry_arn).map(&:to_h)
  expected_service_registries = @service_discovery&.service_registries&.sort_by { |s| s[:registry_arn] } || []
  if actual_service_registries != expected_service_registries
    Hako.logger.warn "Ignoring updated service_registries in the configuration, because AWS doesn't allow updating them for now."
  end
end