Class: Cnvrg::Helpers::Agent

Inherits:
Object
  • Object
show all
Defined in:
lib/cnvrg/helpers/agent.rb

Defined Under Namespace

Modules: LogLevel, Status

Instance Method Summary collapse

Constructor Details

#initialize(executer: nil, slug: nil, command: nil, container_name: nil, send_log_interval: 60, timeout: -1,, logs_regex: [], async: false, send_logs: false, files_exist: [], retries: 0, sleep_before_retry: 30, single_quotes: false, docker_user: nil, use_bash: false, **kwargs) ⇒ Agent

This class represent a single command in the system. it runs under an executer (machine_activity) so it should have all the executer params



19
20
21
22
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
 
# File 'lib/cnvrg/helpers/agent.rb', line 19

def initialize(executer: nil, slug: nil, command: nil, container_name: nil, send_log_interval: 60, timeout: -1, logs_regex: [], async: false, send_logs: false, files_exist: [], retries: 0, sleep_before_retry: 30, single_quotes: false, docker_user: nil, use_bash: false, **kwargs)
  @executer = executer
  @slug = slug
  @files_exist = files_exist
  @container_name = container_name
  @run_in_slave = @container_name.downcase == "slave"
  @log_interval = send_log_interval
  # https://ruby-doc.org/stdlib-2.5.1/libdoc/timeout/rdoc/Timeout.html timeout should be 0 for running forever
  if timeout.blank? or timeout.negative?
    @timeout = 0
  else
    @timeout = timeout
  end
  @logs_regex = logs_regex || []
  @async = async
  @command = command
  @send_logs = send_logs
  @retries = retries.try(:to_i) ## How many times the user asked to try to execute the command again
  @sleep_before_retry = sleep_before_retry
  @real_execution_retries = 0 ## How many times the command really executed until success
  @single_quotes = single_quotes
  @docker_user = ""
  @shell_type = use_bash ? "bash -l" : "sh"
  if docker_user.present?
    @docker_user = " --user #{docker_user}"
  end
  if @run_in_slave
    if @single_quotes
      @command = "docker exec #{@docker_user} -it #{@executer.slave_id} #{@shell_type} -c '#{@command}'"
    else
      @command = "docker exec #{@docker_user} -it #{@executer.slave_id} #{@shell_type} -c \"#{@command}\""
    end
  end
  @output = []
  @errors = []
  @exit_status = nil
  @is_running = true
  @pid = nil
end

Instance Method Details

#base_url ⇒ Object 



59
60
61
 
# File 'lib/cnvrg/helpers/agent.rb', line 59

def base_url
  [@executer.activity_url, "commands", @slug].join("/")
end

#exec! ⇒ Object 



76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
 
# File 'lib/cnvrg/helpers/agent.rb', line 76

def exec!
  log_internal("Command: #{@command} with slug: #{@slug} started!")
  if @command.blank?
    @exit_status = 0
  elsif should_run?
    send_logs(status: Status::STARTED)
    periodic_thread_handle = periodic_thread
    execute_command
  else
    @exit_status = 127
  end
  finish_log = "Command: #{@command} with slug: #{@slug} finished"
  finish_log += " after #{@real_execution_retries} retries" if @real_execution_retries > 0
  log_internal(finish_log)
  send_logs(exit_status: @exit_status, status: Status::FINISHED)
  if periodic_thread_handle.present?
    periodic_thread_handle.join
  end
end

#execute_command ⇒ Object 



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
 
# File 'lib/cnvrg/helpers/agent.rb', line 120

def execute_command
  Timeout.timeout(@timeout) do
    PTY.spawn(@command) do |stdout, stdin, pid, stderr|
      @pid = pid
      begin
        if stdout.present?
          stdout.each do |line|
            log_internal(line, level: LogLevel::PURE)
            line = line.strip.gsub(/\e\[([;\d]+)?m/, '')
            @output << {log: line, timestamp: Time.now}
          end
        end

        if stderr.present?
          stderr.each do |line|
            line = line.strip.gsub(/\e\[([;\d]+)?m/, '')
            log_internal(line, level: LogLevel::ERROR)
            @errors << {log: line, timestamp: Time.now}
          end
        end
      rescue Errno::EIO => e
        next
      rescue => e
        log_internal(e.message, level: LogLevel::ERROR)
        log_internal(e.backtrace.join("\n"), level: LogLevel::ERROR)
        @errors << {log: e.message, timestamp: Time.now}
      end
      ::Process.wait pid
    end
  end
  @exit_status = $?.exitstatus
rescue Timeout::Error
  Process.kill(0, @pid)
  @errors << {log: "Command timed out!", timestamp: Time.now}
  log_internal("Command timed out!", level: LogLevel::ERROR)
  @exit_status = 124
ensure
  retry_command if @retries != 0 and @exit_status !=0
  @exit_status
end

#get_logs_to_send ⇒ Object 



96
97
98
99
100
 
# File 'lib/cnvrg/helpers/agent.rb', line 96

def get_logs_to_send
  new_logs = @output.pop(@output.length)
  new_errors = @errors.pop(@errors.length)
  [new_logs, new_errors]
end

#periodic_thread ⇒ Object 



103
104
105
106
107
108
109
110
111
 
# File 'lib/cnvrg/helpers/agent.rb', line 103

def periodic_thread
  Thread.new do
      while @exit_status.blank?
        Thread.exit if @log_interval.blank?
        sleep(@log_interval)
        send_logs
      end
  end
end

#retry_command ⇒ Object 



113
114
115
116
117
118
 
# File 'lib/cnvrg/helpers/agent.rb', line 113

def retry_command
  @retries -=1
  sleep @sleep_before_retry
  @real_execution_retries +=1
  execute_command
end

#should_run? ⇒ Boolean

Returns:

  • (Boolean) 


63
64
65
66
67
68
69
70
71
72
73
 
# File 'lib/cnvrg/helpers/agent.rb', line 63

def should_run?
  if @files_exist.present?
    file_doesnt_exists = @files_exist.find do |file|
      not File.exists? file
    end
    return true if file_doesnt_exists.blank?
    log_internal("Can't find file #{file_doesnt_exists}, stopping the job")
    return false
  end
  true
end