Class: Bosh::Director::AgentClient

Inherits:
Object
  • Object
show all
Defined in:
lib/bosh/director/agent_client.rb

Constant Summary collapse

DEFAULT_POLL_INTERVAL = 
1.0
GET_TASK_MAX_RETRIES =

in case of timeout errors

2
GET_STATE_MAX_RETRIES =

get_task should retry at least once because some long running tasks (e.g. configure_networks) will restart the agent (current implementation) which most likely will result in first get_task message being lost because agent was not listening on NATS and second retry message will probably be received because agent came back up.

2

Instance Attribute Summary collapse

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(service_name, client_id, options = {}) ⇒ AgentClient

Returns a new instance of AgentClient.



38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
 
# File 'lib/bosh/director/agent_client.rb', line 38

def initialize(service_name, client_id, options = {})
  @service_name = service_name
  @client_id = client_id
  @nats_rpc = Config.nats_rpc
  @timeout = options[:timeout] || 45
  @logger = Config.logger
  @retry_methods = options[:retry_methods] || {}

  if options[:credentials]
    @encryption_handler =
      Bosh::Core::EncryptionHandler.new(@client_id, options[:credentials])
  end

  @resource_manager = Api::ResourceManager.new
end

Dynamic Method Handling

This class handles dynamic methods through the method_missing method

#method_missing(method_name, *args) ⇒ Object 



54
55
56
 
# File 'lib/bosh/director/agent_client.rb', line 54

def method_missing(method_name, *args)
  send_message(method_name, *args)
end

Instance Attribute Details

#id ⇒ Object

Returns the value of attribute id.



17
18
19
 
# File 'lib/bosh/director/agent_client.rb', line 17

def id
  @id
end

Class Method Details

.with_defaults(id, options = {}) ⇒ Object 



19
20
21
22
 
# File 'lib/bosh/director/agent_client.rb', line 19

def self.with_defaults(id, options = {})
  vm = Bosh::Director::Models::Vm.find(:agent_id => id)
  with_vm(vm, options)
end

.with_vm(vm, options = {}) ⇒ Object 



24
25
26
27
28
29
30
31
32
33
34
35
36
 
# File 'lib/bosh/director/agent_client.rb', line 24

def self.with_vm(vm, options = {})
  defaults = {
    retry_methods: {
      get_state: GET_STATE_MAX_RETRIES,
      get_task: GET_TASK_MAX_RETRIES,
    }
  }

  credentials = vm.credentials
  defaults.merge!(credentials: credentials) if credentials

  self.new('agent', vm.agent_id, defaults.merge(options))
end

Instance Method Details

#apply(*args) ⇒ Object 



77
78
79
 
# File 'lib/bosh/director/agent_client.rb', line 77

def apply(*args)
  send_long_running_message(:apply, *args)
end

#compile_package(*args) ⇒ Object 



81
82
83
 
# File 'lib/bosh/director/agent_client.rb', line 81

def compile_package(*args)
  send_long_running_message(:compile_package, *args)
end

#configure_networks(*args) ⇒ Object 



137
138
139
 
# File 'lib/bosh/director/agent_client.rb', line 137

def configure_networks(*args)
  send_long_running_message(:configure_networks, *args)
end

#drain(*args) ⇒ Object 



85
86
87
 
# File 'lib/bosh/director/agent_client.rb', line 85

def drain(*args)
  send_long_running_message(:drain, *args)
end

#fetch_logs(*args) ⇒ Object 



89
90
91
 
# File 'lib/bosh/director/agent_client.rb', line 89

def fetch_logs(*args)
  send_long_running_message(:fetch_logs, *args)
end

#format_exception(exception) ⇒ String

Returns formatted exception information

Parameters:

  • exception (Hash|#to_s) —

    Serialized exception

Returns:

  • (String) 


216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
 
# File 'lib/bosh/director/agent_client.rb', line 216

def format_exception(exception)
  return exception.to_s unless exception.is_a?(Hash)

  msg = exception["message"].to_s

  if exception["backtrace"]
    msg += "\n"
    msg += Array(exception["backtrace"]).join("\n")
  end

  if exception["blobstore_id"]
    blob = download_and_delete_blob(exception["blobstore_id"])
    msg += "\n"
    msg += blob.to_s
  end

  msg
end

#handle_method(method_name, args) ⇒ Object 



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
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
 
# File 'lib/bosh/director/agent_client.rb', line 159

def handle_method(method_name, args)
  result = {}
  result.extend(MonitorMixin)

  cond = result.new_cond
  timeout_time = Time.now.to_f + @timeout

  request = { :method => method_name, :arguments => args }

  if @encryption_handler
    @logger.info("Request: #{request}")
    request = { "encrypted_data" => @encryption_handler.encrypt(request) }
    request["session_id"] = @encryption_handler.session_id
  end

  recipient = "#{@service_name}.#{@client_id}"

  request_id = @nats_rpc.send_request(recipient, request) do |response|
    if @encryption_handler
      begin
        response = @encryption_handler.decrypt(response["encrypted_data"])
      rescue Bosh::Core::EncryptionHandler::CryptError => e
        response["exception"] = "CryptError: #{e.inspect} #{e.backtrace}"
      end
      @logger.info("Response: #{response}")
    end

    result.synchronize do
      inject_compile_log(response)
      result.merge!(response)
      cond.signal
    end
  end

  result.synchronize do
    while result.empty?
      timeout = timeout_time - Time.now.to_f
      unless timeout > 0
        @nats_rpc.cancel_request(request_id)
        raise RpcTimeout,
              "Timed out sending `#{method_name}' to #{@client_id} " +
                "after #{@timeout} seconds"
      end
      cond.wait(timeout)
    end
  end

  if result.has_key?("exception")
    raise RpcRemoteException, format_exception(result["exception"])
  end

  result["value"]
end

#migrate_disk(*args) ⇒ Object 



93
94
95
 
# File 'lib/bosh/director/agent_client.rb', line 93

def migrate_disk(*args)
  send_long_running_message(:migrate_disk, *args)
end

#mount_disk(*args) ⇒ Object 



97
98
99
 
# File 'lib/bosh/director/agent_client.rb', line 97

def mount_disk(*args)
  send_long_running_message(:mount_disk, *args)
end

#prepare(*args) ⇒ Object 



73
74
75
 
# File 'lib/bosh/director/agent_client.rb', line 73

def prepare(*args)
  send_long_running_message(:prepare, *args)
end

#start_errand(*args) ⇒ Object 



121
122
123
 
# File 'lib/bosh/director/agent_client.rb', line 121

def start_errand(*args)
  start_long_running_task(:run_errand, *args)
end

#stop(*args) ⇒ Object 



117
118
119
 
# File 'lib/bosh/director/agent_client.rb', line 117

def stop(*args)
  send_long_running_message(:stop, *args)
end

#unmount_disk(*args) ⇒ Object 



101
102
103
 
# File 'lib/bosh/director/agent_client.rb', line 101

def unmount_disk(*args)
  send_long_running_message(:unmount_disk, *args)
end

#update_settings(certs) ⇒ Object 



105
106
107
108
109
110
111
112
113
114
115
 
# File 'lib/bosh/director/agent_client.rb', line 105

def update_settings(certs)
  begin
    send_long_running_message(:update_settings, {"trusted_certs" => certs})
  rescue RpcRemoteException => e
    if e.message =~ /unknown message/
      @logger.warn("Ignoring update_settings 'unknown message' error from the agent: #{e.inspect}")
    else
      raise
    end
  end
end

#wait_for_task(agent_task_id, &blk) ⇒ Object 



125
126
127
128
129
130
131
132
133
134
135
 
# File 'lib/bosh/director/agent_client.rb', line 125

def wait_for_task(agent_task_id, &blk)
  task = get_task_status(agent_task_id)

  while task['state'] == 'running'
    blk.call if block_given?
    sleep(DEFAULT_POLL_INTERVAL)
    task = get_task_status(agent_task_id)
  end

  task['value']
end

#wait_until_ready(deadline = 600) ⇒ Object 



141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
 
# File 'lib/bosh/director/agent_client.rb', line 141

def wait_until_ready(deadline = 600)
  old_timeout = @timeout
  @timeout = 1.0
  @deadline = Time.now.to_i + deadline

  begin
    ping
  rescue RpcTimeout
    retry if @deadline - Time.now.to_i > 0
    raise RpcTimeout, "Timed out pinging to #{@client_id} after #{deadline} seconds"
  rescue RpcRemoteException => e
    retry if e.message =~ /^restarting agent/ && @deadline - Time.now.to_i > 0
    raise e
  ensure
    @timeout = old_timeout
  end
end