Class: Bosh::AwsCloud::Cloud

Inherits:
Cloud
  • Object
show all
Includes:
Helpers
Defined in:
lib/cloud/aws/cloud.rb

Constant Summary collapse

DEFAULT_MAX_RETRIES =

default maximum number of times to retry an AWS API call

2
METADATA_TIMEOUT =

in seconds

5
DEVICE_POLL_TIMEOUT =

in seconds

60

Instance Attribute Summary collapse

Instance Method Summary collapse

Methods included from Helpers

#cloud_error, #default_ephemeral_disk_mapping, #ebs_ephemeral_disk_mapping, #extract_security_groups

Constructor Details

#initialize(options) ⇒ Cloud

Initialize BOSH AWS CPI. The contents of sub-hashes are defined in the README

Parameters:

  • options (Hash) —

    CPI options

Options Hash (options):

  • aws (Hash) —

    AWS specific options

  • agent (Hash) —

    agent options

  • registry (Hash) —

    agent options



23
24
25
26
27
28
29
30
31
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
 
# File 'lib/cloud/aws/cloud.rb', line 23

def initialize(options)
  @options = options.dup.freeze
  validate_options
  validate_credentials_source

  @logger = Bosh::Clouds::Config.logger
  aws_logger = @logger

  @aws_params = {
    credentials_source: aws_properties['credentials_source'] || 'static',
    region:            aws_properties['region'],
    max_retries:       aws_properties['max_retries']  || DEFAULT_MAX_RETRIES,
    logger:            aws_logger
  }

  %w(
    http_read_timeout
    http_wire_trace
    proxy_uri
    ssl_verify_peer
    ssl_ca_file
    ssl_ca_path
  ).each do |k|
    @aws_params[k.to_sym] = aws_properties[k] unless aws_properties[k].nil?
  end

  # credentials_source could be static (default) or env_or_profile
  # static credentials must be included in aws_properties
  # env_or_profile credentials will use the AWS DefaultCredentialsProvider
  # to find AWS credentials in environment variables or EC2 instance profiles

  if @aws_params[:credentials_source] == 'static'
    @aws_params[:access_key_id] = aws_properties['access_key_id']
    @aws_params[:secret_access_key] = aws_properties['secret_access_key']
  end

  # AWS Ruby SDK is threadsafe but Ruby autoload isn't,
  # so we need to trigger eager autoload while constructing CPI
  AWS.eager_autoload!

  initialize_aws
  initialize_registry

  elb = AWS::ELB.new(@aws_params)

  @instance_manager = InstanceManager.new(region, registry, elb, az_selector, @logger)

  @metadata_lock = Mutex.new
end

Instance Attribute Details

#ec2 ⇒ Object (readonly)

Returns the value of attribute ec2.



12
13
14
 
# File 'lib/cloud/aws/cloud.rb', line 12

def ec2
  @ec2
end

#logger ⇒ Object

Returns the value of attribute logger.



15
16
17
 
# File 'lib/cloud/aws/cloud.rb', line 15

def logger
  @logger
end

#options ⇒ Object (readonly)

Returns the value of attribute options.



14
15
16
 
# File 'lib/cloud/aws/cloud.rb', line 14

def options
  @options
end

#registry ⇒ Object (readonly)

Returns the value of attribute registry.



13
14
15
 
# File 'lib/cloud/aws/cloud.rb', line 13

def registry
  @registry
end

Instance Method Details

#attach_disk(instance_id, disk_id) ⇒ Object

Attach an EBS volume to an EC2 instance

Parameters:

  • instance_id (String) —

    EC2 instance id of the virtual machine to attach the disk to

  • disk_id (String) —

    EBS volume id of the disk to attach



279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
 
# File 'lib/cloud/aws/cloud.rb', line 279

def attach_disk(instance_id, disk_id)
  with_thread_name("attach_disk(#{instance_id}, #{disk_id})") do
    instance = @ec2.instances[instance_id]
    volume = @ec2.volumes[disk_id]

    device_name = attach_ebs_volume(instance, volume)

    update_agent_settings(instance) do |settings|
      settings["disks"] ||= {}
      settings["disks"]["persistent"] ||= {}
      settings["disks"]["persistent"][disk_id] = device_name
    end
    logger.info("Attached `#{disk_id}' to `#{instance_id}'")
  end

  # log registry settings for debugging
  logger.debug("updated registry settings: #{registry.read_settings(instance_id)}")
end

#compare_private_ip_addresses(instance, specified_ip_address) ⇒ void

This method returns an undefined value.

Compares actual instance private IP addresses with the IP address specified at the network spec

change the IP address of a running server, so we need to send the InstanceUpdater a request to do it for us

Parameters:

  • instance (AWS::EC2::Instance) —

    EC2 instance

  • specified_ip_address (String) —

    IP address specified at the network spec (if Manual Network)

Raises:

  • (Bosh::Clouds:NotSupported) —

    If the IP address change, we need to recreate the VM as you can’t



426
427
428
429
430
431
432
433
434
 
# File 'lib/cloud/aws/cloud.rb', line 426

def compare_private_ip_addresses(instance, specified_ip_address)
  actual_ip_address = instance.private_ip_address

  unless specified_ip_address.nil? || actual_ip_address == specified_ip_address
    raise Bosh::Clouds::NotSupported,
          "IP address change requires VM recreation: %s to %s" %
          [actual_ip_address, specified_ip_address]
  end
end

#compare_security_groups(instance, network_spec) ⇒ Object

If the security groups change, we need to recreate the VM as you can’t change the security group of a running instance, we need to send the InstanceUpdater a request to do it for us



402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
 
# File 'lib/cloud/aws/cloud.rb', line 402

def compare_security_groups(instance, network_spec)
  actual_group_names = instance.security_groups.collect { |sg| sg.name }
  specified_group_names = extract_security_groups(network_spec)
  if specified_group_names.empty?
    new_group_names = Array(aws_properties["default_security_groups"])
  else
    new_group_names = specified_group_names
  end

  unless actual_group_names.sort == new_group_names.sort
    raise Bosh::Clouds::NotSupported,
          "security groups change requires VM recreation: %s to %s" %
              [actual_group_names.join(", "), new_group_names.join(", ")]
  end
end

#configure_networks(instance_id, network_spec) ⇒ Object

Configure network for an EC2 instance

Parameters:

  • instance_id (String) —

    EC2 instance id

  • network_spec (Hash) —

    network properties

Raises:

  • (Bosh::Clouds:NotSupported) —

    if there’s a network change that requires the recreation of the VM



379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
 
# File 'lib/cloud/aws/cloud.rb', line 379

def configure_networks(instance_id, network_spec)
  with_thread_name("configure_networks(#{instance_id}, ...)") do
    logger.info("Configuring '#{instance_id}' to use new network settings: #{network_spec.pretty_inspect}")

    instance = @ec2.instances[instance_id]

    network_configurator = NetworkConfigurator.new(network_spec)

    compare_security_groups(instance, network_spec)

    compare_private_ip_addresses(instance, network_configurator.private_ip)

    network_configurator.configure(@ec2, instance)

    update_agent_settings(instance) do |settings|
      settings["networks"] = agent_network_spec(network_spec)
    end
  end
end

#create_disk(size, cloud_properties, instance_id = nil) ⇒ String

Creates a new EBS volume

Parameters:

  • size (Integer) —

    disk size in MiB

  • instance_id (optional, String) (defaults to: nil) —

    EC2 instance id of the VM that this disk will be attached to

Returns:

  • (String) —

    created EBS volume id

Raises:

  • (ArgumentError) 


190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
 
# File 'lib/cloud/aws/cloud.rb', line 190

def create_disk(size, cloud_properties, instance_id = nil)
  raise ArgumentError, 'disk size needs to be an integer' unless size.kind_of?(Integer)
  with_thread_name("create_disk(#{size}, #{instance_id})") do
    volume_properties = VolumeProperties.new(
      size: size,
      type: cloud_properties['type'],
      iops: cloud_properties['iops'],
      az: @az_selector.select_availability_zone(instance_id),
      encrypted: cloud_properties['encrypted']
    )
    volume_properties.validate!

    # if the disk is created for an instance, use the same availability zone as they must match
    with_volume_options = VolumesCreatePresenter.new(volume_properties).present
    volume = @ec2.volumes.create(with_volume_options)

    logger.info("Creating volume '#{volume.id}'")
    ResourceWait.for_volume(volume: volume, state: :available)

    volume.id
  end
end

#create_stemcell(image_path, stemcell_properties) ⇒ String

Creates a new EC2 AMI using stemcell image. This method can only be run on an EC2 instance, as image creation involves creating and mounting new EBS volume as local block device.

Parameters:

  • image_path (String) —

    local filesystem path to a stemcell image

  • cloud_properties (Hash) —

    AWS-specific stemcell properties

Returns:

  • (String) —

    EC2 AMI name of the stemcell



452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
 
# File 'lib/cloud/aws/cloud.rb', line 452

def create_stemcell(image_path, stemcell_properties)
  with_thread_name("create_stemcell(#{image_path}...)") do
    creator = StemcellCreator.new(region, stemcell_properties)

    return creator.fake.id if creator.fake?

    begin
      # These variables are used in 'ensure' clause
      instance = nil
      volume = nil

      # 1. Create and mount new EBS volume (2GB default)
      disk_size = stemcell_properties["disk"] || 2048
      volume_id = create_disk(disk_size, {}, current_vm_id)
      volume = @ec2.volumes[volume_id]
      instance = @ec2.instances[current_vm_id]

      sd_name = attach_ebs_volume(instance, volume)
      ebs_volume = find_ebs_device(sd_name)

      logger.info("Creating stemcell with: '#{volume.id}' and '#{stemcell_properties.inspect}'")
      creator.create(volume, ebs_volume, image_path).id
    rescue => e
      logger.error(e)
      raise e
    ensure
      if instance && volume
        detach_ebs_volume(instance, volume, true)
        delete_disk(volume.id)
      end
    end
  end
end

#create_vm(agent_id, stemcell_id, resource_pool, network_spec, disk_locality = nil, environment = nil) ⇒ String

Create an EC2 instance and wait until it’s in running state

Parameters:

  • agent_id (String) —

    agent id associated with new VM

  • stemcell_id (String) —

    AMI id of the stemcell used to create the new instance

  • resource_pool (Hash) —

    resource pool specification

  • network_spec (Hash) —

    network specification, if it contains security groups they must already exist

  • disk_locality (optional, Array) (defaults to: nil) —

    list of disks that might be attached to this instance in the future, can be used as a placement hint (i.e. instance will only be created if resource pool availability zone is the same as disk availability zone)

  • environment (optional, Hash) (defaults to: nil) —

    data to be merged into agent settings

Returns:

  • (String) —

    EC2 instance id of the new virtual machine



117
118
119
120
121
122
123
124
125
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
 
# File 'lib/cloud/aws/cloud.rb', line 117

def create_vm(agent_id, stemcell_id, resource_pool, network_spec, disk_locality = nil, environment = nil)
  with_thread_name("create_vm(#{agent_id}, ...)") do
    # do this early to fail fast
    stemcell = StemcellFinder.find_by_region_and_id(region, stemcell_id)

    begin
      instance, block_device_agent_info = @instance_manager.create(
        agent_id,
        stemcell.image_id,
        resource_pool,
        network_spec,
        (disk_locality || []),
        environment,
        options,
      )

      logger.info("Creating new instance '#{instance.id}'")

      NetworkConfigurator.new(network_spec).configure(region, instance)

      registry_settings = initial_agent_settings(
        agent_id,
        network_spec,
        environment,
        stemcell.root_device_name,
        block_device_agent_info
      )
      registry.update_settings(instance.id, registry_settings)

      instance.id
    rescue => e # is this rescuing too much?
      logger.error(%Q[Failed to create instance: #{e.message}\n#{e.backtrace.join("\n")}])
      instance.terminate(fast_path_delete?) if instance
      raise e
    end
  end
end

#current_vm_id ⇒ Object

Reads current instance id from EC2 metadata. We are assuming instance id cannot change while current process is running and thus memoizing it.



77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
 
# File 'lib/cloud/aws/cloud.rb', line 77

def current_vm_id
  @metadata_lock.synchronize do
    return @current_vm_id if @current_vm_id

    client = HTTPClient.new
    client.connect_timeout = METADATA_TIMEOUT
    # Using 169.254.169.254 is an EC2 convention for getting
    # instance metadata
    uri = "http://169.254.169.254/latest/meta-data/instance-id/"

    response = client.get(uri)
    unless response.status == 200
      cloud_error("Instance metadata endpoint returned " \
                  "HTTP #{response.status}")
    end

    @current_vm_id = response.body
  end

rescue HTTPClient::TimeoutError
  cloud_error("Timed out reading instance metadata, " \
              "please make sure CPI is running on EC2 instance")
end

#delete_disk(disk_id) ⇒ Object

Delete EBS volume

Parameters:

  • disk_id (String) —

    EBS volume id

Raises:

  • (Bosh::Clouds::CloudError) —

    if disk is not in available state



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
262
263
264
265
266
267
268
269
270
271
272
273
274
 
# File 'lib/cloud/aws/cloud.rb', line 229

def delete_disk(disk_id)
  with_thread_name("delete_disk(#{disk_id})") do
    volume = @ec2.volumes[disk_id]

    logger.info("Deleting volume `#{volume.id}'")

    # Retry 1, 6, 11, 15, 15, 15.. seconds. The total time is ~10 min.
    # VolumeInUse can be returned by AWS if disk was attached to VM
    # that was recently removed.
    tries = ResourceWait::DEFAULT_WAIT_ATTEMPTS
    sleep_cb = ResourceWait.sleep_callback(
      "Waiting for volume `#{volume.id}' to be deleted",
      { interval: 5, total: tries }
    )
    ensure_cb = Proc.new do |retries|
      cloud_error("Timed out waiting to delete volume `#{volume.id}'") if retries == tries
    end
    errors = [AWS::EC2::Errors::VolumeInUse, AWS::EC2::Errors::RequestLimitExceeded]

    Bosh::Common.retryable(tries: tries, sleep: sleep_cb, on: errors, ensure: ensure_cb) do
      begin
        volume.delete
      rescue AWS::EC2::Errors::InvalidVolume::NotFound => e
        logger.warn("Failed to delete disk '#{disk_id}' because it was not found: #{e.inspect}")
        raise Bosh::Clouds::DiskNotFound.new(false), "Disk '#{disk_id}' not found"
      end

      true # return true to only retry on Exceptions
    end

    if fast_path_delete?
      begin
        TagManager.tag(volume, "Name", "to be deleted")
        logger.info("Volume `#{disk_id}' has been marked for deletion")
      rescue AWS::EC2::Errors::InvalidVolume::NotFound
        # Once in a blue moon AWS if actually fast enough that the volume is already gone
        # when we get here, and if it is, our work here is done!
      end
      return
    end

    ResourceWait.for_volume(volume: volume, state: :deleted)

    logger.info("Volume `#{disk_id}' has been deleted")
  end
end

#delete_snapshot(snapshot_id) ⇒ Object

Delete a disk snapshot

Parameters:

  • snapshot_id (String) —

    snapshot id to delete



362
363
364
365
366
367
368
369
370
371
372
373
 
# File 'lib/cloud/aws/cloud.rb', line 362

def delete_snapshot(snapshot_id)
  with_thread_name("delete_snapshot(#{snapshot_id})") do
    snapshot = @ec2.snapshots[snapshot_id]

    if snapshot.status == :in_use
      raise Bosh::Clouds::CloudError, "snapshot '#{snapshot.id}' can not be deleted as it is in use"
    end

    snapshot.delete
    logger.info("snapshot '#{snapshot_id}' deleted")
  end
end

#delete_stemcell(stemcell_id) ⇒ Object

Delete a stemcell and the accompanying snapshots

Parameters:

  • stemcell_id (String) —

    EC2 AMI name of the stemcell to be deleted



488
489
490
491
492
493
 
# File 'lib/cloud/aws/cloud.rb', line 488

def delete_stemcell(stemcell_id)
  with_thread_name("delete_stemcell(#{stemcell_id})") do
    stemcell = StemcellFinder.find_by_region_and_id(region, stemcell_id)
    stemcell.delete
  end
end

#delete_vm(instance_id) ⇒ Object

Delete EC2 instance (“terminate” in AWS language) and wait until it reports as terminated

Parameters:

  • instance_id (String) —

    EC2 instance id



159
160
161
162
163
164
 
# File 'lib/cloud/aws/cloud.rb', line 159

def delete_vm(instance_id)
  with_thread_name("delete_vm(#{instance_id})") do
    logger.info("Deleting instance '#{instance_id}'")
    @instance_manager.find(instance_id).terminate(fast_path_delete?)
  end
end

#detach_disk(instance_id, disk_id) ⇒ Object

Detach an EBS volume from an EC2 instance

Parameters:

  • instance_id (String) —

    EC2 instance id of the virtual machine to detach the disk from

  • disk_id (String) —

    EBS volume id of the disk to detach



301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
 
# File 'lib/cloud/aws/cloud.rb', line 301

def detach_disk(instance_id, disk_id)
  with_thread_name("detach_disk(#{instance_id}, #{disk_id})") do
    instance = @ec2.instances[instance_id]
    volume = @ec2.volumes[disk_id]

    if volume.exists?
      detach_ebs_volume(instance, volume)
    else
      @logger.info("Disk `#{disk_id}' not found while trying to detach it from vm `#{instance_id}'...")
    end

    update_agent_settings(instance) do |settings|
      settings["disks"] ||= {}
      settings["disks"]["persistent"] ||= {}
      settings["disks"]["persistent"].delete(disk_id)
    end

    logger.info("Detached `#{disk_id}' from `#{instance_id}'")
  end
end

#find_ebs_device(sd_name) ⇒ Object 



528
529
530
531
532
533
534
535
536
537
538
539
540
541
 
# File 'lib/cloud/aws/cloud.rb', line 528

def find_ebs_device(sd_name)
  xvd_name = sd_name.gsub(/^\/dev\/sd/, "/dev/xvd")

  DEVICE_POLL_TIMEOUT.times do
    if File.blockdev?(sd_name)
      return sd_name
    elsif File.blockdev?(xvd_name)
      return xvd_name
    end
    sleep(1)
  end

  cloud_error("Cannot find EBS volume on current instance")
end

#get_disks(vm_id) ⇒ Object 



322
323
324
325
326
327
328
329
330
 
# File 'lib/cloud/aws/cloud.rb', line 322

def get_disks(vm_id)
  disks = []
  @ec2.instances[vm_id].block_devices.each do |block_device|
    if block_device[:ebs]
      disks << block_device[:ebs][:volume_id]
    end
  end
  disks
end

#has_disk?(disk_id) ⇒ bool

Check whether an EBS volume exists or not

Parameters:

  • disk_id (String) —

    EBS volume UUID

Returns:

  • (bool) —

    whether the specific disk is there or not



218
219
220
221
222
223
 
# File 'lib/cloud/aws/cloud.rb', line 218

def has_disk?(disk_id)
  with_thread_name("has_disk?(#{disk_id})") do
    @logger.info("Check the presence of disk with id `#{disk_id}'...")
    @ec2.volumes[disk_id].exists?
  end
end

#has_vm?(instance_id) ⇒ Boolean

Has EC2 instance

Parameters:

  • instance_id (String) —

    EC2 instance id

Returns:

  • (Boolean) 


178
179
180
181
182
 
# File 'lib/cloud/aws/cloud.rb', line 178

def has_vm?(instance_id)
  with_thread_name("has_vm?(#{instance_id})") do
    @instance_manager.find(instance_id).exists?
  end
end

#reboot_vm(instance_id) ⇒ Object

Reboot EC2 instance

Parameters:

  • instance_id (String) —

    EC2 instance id



169
170
171
172
173
 
# File 'lib/cloud/aws/cloud.rb', line 169

def reboot_vm(instance_id)
  with_thread_name("reboot_vm(#{instance_id})") do
    @instance_manager.find(instance_id).reboot
  end
end

#set_vm_metadata(vm, metadata) ⇒ void

This method returns an undefined value.

Add tags to an instance. In addition to the suplied tags, it adds a ‘Name’ tag as it is shown in the AWS console.

Parameters:

  • vm (String) —

    vm id that was once returned by #create_vm

  • metadata (Hash) —

    metadata key/value pairs



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
 
# File 'lib/cloud/aws/cloud.rb', line 500

def set_vm_metadata(vm, metadata)
  metadata = Hash[metadata.map{|key,value| [key.to_s, value] }]

  instance = @ec2.instances[vm]

  metadata.each_pair do |key, value|
    TagManager.tag(instance, key, value) unless key == 'name'
  end

  name = metadata['name']
  if name
    TagManager.tag(instance, "Name", name)
    return
  end

  job = metadata['job']
  index = metadata['index']

  if job && index
    name = "#{job}/#{index}"
  elsif metadata['compiling']
    name = "compiling/#{metadata['compiling']}"
  end
  TagManager.tag(instance, "Name", name) if name
rescue AWS::EC2::Errors::TagLimitExceeded => e
  logger.error("could not tag #{instance.id}: #{e.message}")
end

#snapshot_disk(disk_id, metadata) ⇒ String

Take snapshot of disk

Parameters:

  • disk_id (String) —

    disk id of the disk to take the snapshot of

Returns:

  • (String) —

    snapshot id



335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
 
# File 'lib/cloud/aws/cloud.rb', line 335

def snapshot_disk(disk_id, metadata)
  metadata = Hash[metadata.map{|key,value| [key.to_s, value] }]

  with_thread_name("snapshot_disk(#{disk_id})") do
    volume = @ec2.volumes[disk_id]
    devices = []
    volume.attachments.each {|attachment| devices << attachment.device}

    name = ['deployment', 'job', 'index'].collect { |key| metadata[key] }
    name << devices.first.split('/').last unless devices.empty?

    snapshot = volume.create_snapshot(name.join('/'))
    logger.info("snapshot '#{snapshot.id}' of volume '#{disk_id}' created")

    ['agent_id', 'instance_id', 'director_name', 'director_uuid'].each do |key|
      TagManager.tag(snapshot, key, metadata[key])
    end
    TagManager.tag(snapshot, 'device', devices.first) unless devices.empty?
    TagManager.tag(snapshot, 'Name', name.join('/'))

    ResourceWait.for_snapshot(snapshot: snapshot, state: :completed)
    snapshot.id
  end
end