Class: Aspera::Agent::Direct

Inherits:
Base
  • Object
show all
Defined in:
lib/aspera/agent/direct.rb

Overview

executes a local “ascp”, create mgt port

Instance Method Summary collapse

Methods inherited from Base

agent_list, factory_create, #wait_for_completion

Constructor Details

#initialize(ascp_args: nil, wss: true, quiet: true, trusted_certs: nil, client_ssh_key: nil, check_ignore_cb: nil, spawn_timeout_sec: 2, spawn_delay_sec: 2, multi_incr_udp: nil, resume: nil, monitor: true, management_cb: nil, **base_options) ⇒ Direct

options for initialize (same as values in option transfer_info)

Parameters:

  • ascp_args (Array) (defaults to: nil) —

    additional arguments to ascp

  • wss (Boolean) (defaults to: true) —

    true: if both SSH and wss in ts: prefer wss

  • quiet (Boolean) (defaults to: true) —

    by default no native ascp progress bar

  • monitor (Boolean) (defaults to: true) —

    set to false to eliminate management port

  • trusted_certs (Array, NilClass) (defaults to: nil) —

    list of files with trusted certificates (stores)

  • client_ssh_key (String) (defaults to: nil) —

    client ssh key option (from CLIENT_SSH_KEY_OPTIONS)

  • check_ignore_cb (Proc) (defaults to: nil) —

    callback with host,port

  • spawn_timeout_sec (Integer) (defaults to: 2) —

    timeout for ascp spawn

  • spawn_delay_sec (Integer) (defaults to: 2) —

    optional delay to start between sessions

  • multi_incr_udp (Boolean, NilClass) (defaults to: nil) —

    true: increment udp port for each session

  • resume (Hash, NilClass) (defaults to: nil) —

    resume policy

  • management_cb (Proc) (defaults to: nil) —

    callback for management events

  • base_options (Hash) —

    other options for base class



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
 
# File 'lib/aspera/agent/direct.rb', line 41

def initialize(
  ascp_args:         nil,
  wss:               true,
  quiet:             true,
  trusted_certs:     nil,
  client_ssh_key:    nil,
  check_ignore_cb:   nil,
  spawn_timeout_sec: 2,
  spawn_delay_sec:   2,
  multi_incr_udp:    nil,
  resume:            nil,
  monitor:           true,
  management_cb:     nil,
  **base_options
)
  super(**base_options)
  # special transfer parameters
  @tr_opts = {
    ascp_args:       ascp_args,
    wss:             wss,
    quiet:           quiet,
    trusted_certs:   trusted_certs,
    client_ssh_key:  client_ssh_key,
    check_ignore_cb: check_ignore_cb
  }
  @spawn_timeout_sec = spawn_timeout_sec
  @spawn_delay_sec = spawn_delay_sec
  # default is true on Windows, false on other OSes
  @multi_incr_udp = multi_incr_udp.nil? ? Environment.os.eql?(Environment::OS_WINDOWS) : multi_incr_udp
  @monitor = monitor
  @management_cb = management_cb
  @resume_policy = Resumer.new(resume.nil? ? {} : resume.symbolize_keys)
  # all transfer jobs, key = SecureRandom.uuid, protected by mutex, cond var on change
  @sessions = []
  # mutex protects global data accessed by threads
  @mutex = Mutex.new
  @pre_calc_sent = false
  @pre_calc_last_size = nil
end

Instance Method Details

#send_command(job_id, data) ⇒ Object

send command to management port of command (used in ‘asession) ’type’=>‘START’,‘source’=><em>path</em>,‘destination’=><em>path</em> ‘type’=>‘DONE’

Parameters:

  • job_id —

    identified transfer process

  • session_index —

    index of session (for multi session)

  • data —

    command on mgt port, examples:



196
197
198
199
200
 
# File 'lib/aspera/agent/direct.rb', line 196

def send_command(job_id, data)
  session = @sessions.find{ |session| session[:job_id].eql?(job_id)}
  Log.log.debug{"command: #{data}"}
  session[:io].puts(Ascp::Management.command_to_stream(data))
end

#sessions_by_job(job_id) ⇒ Array

Returns list of sessions for a job.

Returns:

  • (Array) —

    list of sessions for a job



186
187
188
 
# File 'lib/aspera/agent/direct.rb', line 186

def sessions_by_job(job_id)
  @sessions.select{ |session| session[:job_id].eql?(job_id)}
end

#shutdown ⇒ Object

used by asession (to be removed ?)



181
182
183
 
# File 'lib/aspera/agent/direct.rb', line 181

def shutdown
  Log.log.debug('fasp local shutdown')
end

#start_and_monitor_process(session:, env:, name:, args:) ⇒ nil

This is the low level method to start the transfer process. Typically started in a thread. Start process with management port.

Parameters:

  • session —

    this session information, keys :io and :token_regenerator

  • env (Hash) —

    environment variables (comes from ascp_args)

  • name (Symbol) —

    name of executable: :ascp, :ascp4 or :async (comes from ascp_args)

  • args (Array) —

    command line arguments (comes from ascp_args)

Returns:

  • (nil) —

    when process has exited



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
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
 
# File 'lib/aspera/agent/direct.rb', line 234

def start_and_monitor_process(
  session:,
  env:,
  name:,
  args:
)
  Aspera.assert_type(session, Hash)
  notify_progress(:pre_start, session_id: nil, info: 'starting')
  begin
    command_pid = nil
    command_arguments = []
    if @monitor
      # we use Socket directly, instead of TCPServer, as it gives access to lower level options
      socket_class = defined?(JRUBY_VERSION) ? ServerSocket : Socket
      mgt_server_socket = socket_class.new(Socket::AF_INET, Socket::SOCK_STREAM, 0)
      # open any available (0) local TCP port for use as management port
      mgt_server_socket.bind(Addrinfo.tcp(LISTEN_LOCAL_ADDRESS, SELECT_AVAILABLE_PORT))
      # make port ready to accept connections, before starting ascp
      mgt_server_socket.listen(1)
      # build arguments and add mgt port
      command_arguments = if name.eql?(:async)
        ["--exclusive-mgmt-port=#{mgt_server_socket.local_address.ip_port}"]
      else
        ['-M', mgt_server_socket.local_address.ip_port.to_s]
      end
    end
    command_arguments.concat(args)
    # capture process stderr
    stderr_r, stderr_w = IO.pipe
    # get location of command executable (ascp, async)
    command_path = Ascp::Installation.instance.path(name)
    command_pid = Environment.secure_spawn(env: env, exec: command_path, args: command_arguments, err: stderr_w)
    stderr_w.close
    notify_progress(:pre_start, session_id: nil, info: "waiting for #{name} to start")
    # "ensure" block will wait for process
    return unless @monitor
    # TODO: timeout does not work when Process.spawn is used... until process exits, then it works
    # So we use select to detect that anything happens on the socket (connection)
    Log.log.debug{"before select, timeout: #{@spawn_timeout_sec}"}
    readable, _, _ = IO.select([mgt_server_socket], nil, nil, @spawn_timeout_sec)
    Log.log.debug('after select, before accept')
    Aspera.assert(readable, exception_class: Transfer::Error){'timeout waiting mgt port connect (select not readable)'}
    # There is a connection to accept
    client_socket, _client_addrinfo = mgt_server_socket.accept
    Log.log.debug('after accept')
    management_port_io = client_socket.to_io
    # by default socket is US-ASCII
    # management messages include file names which may be UTF-8
    # TODO: use same value as Encoding.default_external ?
    management_port_io.set_encoding(Encoding::UTF_8)
    session[:io] = management_port_io
    processor = Ascp::Management.new
    # read management port, until socket is closed (gets returns nil)
    while (line = management_port_io.gets)
      event = processor.process_line(line.chomp)
      next unless event
      # event is ready
      Log.log.trace1{Log.dump(:management_port, event)}
      # store session identifier
      session[:id] = event['SessionId'] if event['Type'].eql?('INIT')
      @management_cb&.call(event)
      process_progress(event)
      Log.log.error(event['Description'].to_s) if event['Type'].eql?('FILEERROR') # cspell:disable-line
    end
    Log.log.debug('management io closed')
    # check that last status was received before process exit
    last_event = processor.last_event
    raise Transfer::Error, "No management event (#{last_event.class})" unless last_event.is_a?(Hash)
    case last_event['Type']
    when 'DONE'
      Log.log.trace1{'Graceful shutdown, DONE message received'}
    when 'ERROR'
      if /bearer token/i.match?(last_event['Description']) &&
          session[:token_regenerator].respond_to?(:refreshed_transfer_token)
        # regenerate token here, expired, or error on it
        # Note: in multi-session, each session will have a different one.
        Log.log.warn('Regenerating token for transfer')
        env['ASPERA_SCP_TOKEN'] = session[:token_regenerator].refreshed_transfer_token
      end
      raise Transfer::Error.new(last_event['Description'], last_event['Code'].to_i)
    else
      Log.log.error{"unexpected last event type: #{last_event['Type']}"}
      # raise Transfer::Error, "unexpected last event type: #{last_event['Type']}, #{last_event['Description']}"
    end
  rescue SystemCallError => e
    # Process.spawn failed, or socket error
    raise Transfer::Error, e.message
  rescue Interrupt
    raise Transfer::Error, 'transfer interrupted by user'
  ensure
    mgt_server_socket&.close
    session.delete(:io)
    # if command was successfully started, check its status
    unless command_pid.nil?
      Process.kill(:INT, command_pid) if @monitor && !Environment.os.eql?(Environment::OS_WINDOWS)
      # collect process exit status or wait for termination
      _, status = Process.wait2(command_pid)
      # process stderr of ascp
      stderr_r.each_line do |line|
        Log.log.error(line.chomp)
      end
      stderr_r.close
      # status is nil if an exception occurred before starting command
      if !status&.success?
        message = "#{name} failed (#{status})"
        # raise error only if there was not already an exception (ERROR_INFO)
        raise Transfer::Error, message unless $ERROR_INFO
        # else display this message also, as main exception is already here
        Log.log.error(message)
      end
    end
  end
  nil
end

#start_transfer(transfer_spec, token_regenerator: nil) ⇒ Object

start ascp transfer(s) (non blocking), single or multi-session session information added to @sessions

Parameters:

  • transfer_spec (Hash) —

    aspera transfer specification

  • token_regenerator (Object) (defaults to: nil) —

    object with method refreshed_transfer_token



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
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
154
155
156
157
158
159
160
161
 
# File 'lib/aspera/agent/direct.rb', line 85

def start_transfer(transfer_spec, token_regenerator: nil)
  # clone transfer spec because we modify it (first level keys)
  transfer_spec = transfer_spec.clone
  # if there are aspera tags
  if transfer_spec.dig('tags', Transfer::Spec::TAG_RESERVED).is_a?(Hash)
    # TODO: what is this for ? only on local ascp ?
    # NOTE: important: transfer id must be unique: generate random id
    # using a non unique id results in discard of tags in AoC, and a package is never finalized
    # all sessions in a multi-session transfer must have the same xfer_id (see admin manual)
    transfer_spec['tags'][Transfer::Spec::TAG_RESERVED]['xfer_id'] ||= SecureRandom.uuid
    Log.log.debug{"xfer id=#{transfer_spec['xfer_id']}"}
    # TODO: useful ? node only ? seems to be a timeout for retry in node
    transfer_spec['tags'][Transfer::Spec::TAG_RESERVED]['xfer_retry'] ||= 3600
  end
  Log.log.debug{Log.dump('ts', transfer_spec)}
  # Compute this before using transfer spec because it potentially modifies the transfer spec
  # (even if the var is not used in single session)
  multi_session_info = nil
  if transfer_spec.key?('multi_session')
    # Managed by multi-session, so delete from transfer spec
    multi_session_info = {
      count: transfer_spec.delete('multi_session').to_i
    }
    if multi_session_info[:count].negative?
      Log.log.error{"multi_session(#{transfer_spec['multi_session']}) shall be integer >= 0"}
      multi_session_info = nil
    elsif multi_session_info[:count].eql?(0)
      Log.log.debug('multi_session count is zero: no multi session')
      multi_session_info = nil
    elsif @multi_incr_udp # multi_session_info[:count] > 0
      # if option not true: keep default udp port for all sessions
      multi_session_info[:udp_base] = transfer_spec.key?('fasp_port') ? transfer_spec['fasp_port'] : Transfer::Spec::UDP_PORT
      # delete from original transfer spec, as we will increment values
      transfer_spec.delete('fasp_port')
      # override if specified, else use default value
    end
  end

  # generic session information
  session = {
    id:                nil, # SessionId from INIT message in mgt port
    job_id:            SecureRandom.uuid, # job id (regroup sessions)
    ts:                transfer_spec,     # global transfer spec
    thread:            nil,               # Thread object monitoring management port, not nil when pushed to :sessions
    error:             nil,               # exception if failed
    io:                nil,               # management port server socket
    token_regenerator: token_regenerator, # regenerate bearer token with oauth
    # env vars and args for ascp (from transfer spec)
    env_args:          Transfer::Parameters.new(transfer_spec, **@tr_opts).ascp_args
  }

  if multi_session_info.nil?
    Log.log.debug('Starting single session thread')
    # single session for transfer : simple
    session[:thread] = Thread.new{transfer_thread_entry(session)}
    @sessions.push(session)
  else
    Log.log.debug('Starting multi session threads')
    1.upto(multi_session_info[:count]) do |i|
      # do not delay the first session
      sleep(@spawn_delay_sec) unless i.eql?(1)
      # do deep copy (each thread has its own copy because it is modified here below and in thread)
      this_session = session.clone
      this_session[:ts] = this_session[:ts].clone
      env_args = this_session[:env_args] = this_session[:env_args].clone
      args = env_args[:args] = env_args[:args].clone
      # set multi session part
      args.unshift("-C#{i}:#{multi_session_info[:count]}")
      # option: increment (default as per ascp manual) or not (cluster on other side ?)
      args.unshift('-O', (multi_session_info[:udp_base] + i - 1).to_s) if @multi_incr_udp
      # finally start the thread
      this_session[:thread] = Thread.new{transfer_thread_entry(this_session)}
      @sessions.push(this_session)
    end
  end
  return session[:job_id]
end

#wait_for_transfers_completion ⇒ Object

wait for completion of all jobs started

Returns:

  • list of :success or error message



165
166
167
168
169
170
171
172
173
174
175
176
177
178
 
# File 'lib/aspera/agent/direct.rb', line 165

def wait_for_transfers_completion
  Log.log.debug('wait_for_transfers_completion')
  # set to non-nil to exit loop
  result = []
  @sessions.each do |session|
    Log.log.debug{"join #{session[:thread]}"}
    session[:thread].join
    result.push(session[:error] || :success)
  end
  Log.log.debug('all transfers joined')
  # since all are finished and we return the result, clear statuses
  @sessions.clear
  return result
end