Class: Datadog::Writer

Inherits:
Object
  • Object
show all
Defined in:
lib/ddtrace/writer.rb

Overview

Traces and services writer that periodically sends data to the trace-agent

Constant Summary collapse

HOSTNAME = 
'localhost'.freeze
PORT = 
'8126'.freeze

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(options = {}) ⇒ Writer

Returns a new instance of Writer.



13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
 
# File 'lib/ddtrace/writer.rb', line 13

def initialize(options = {})
  # writer and transport parameters
  @buff_size = options.fetch(:buffer_size, 100)
  @span_interval = options.fetch(:spans_interval, 1)
  @service_interval = options.fetch(:services_interval, 120)

  # transport and buffers
  @transport = options.fetch(:transport, Datadog::HTTPTransport.new(HOSTNAME, PORT))
  @services = {}

  # handles the thread creation after an eventual fork
  @mutex_after_fork = Mutex.new
  @pid = nil

  @traces_flushed = 0
  @services_flushed = 0

  # one worker for both services and traces, each have their own queues
  @worker = nil
end

Instance Attribute Details

#transport ⇒ Object (readonly)

Returns the value of attribute transport.



8
9
10
 
# File 'lib/ddtrace/writer.rb', line 8

def transport
  @transport
end

Instance Method Details

#send_services(services, transport) ⇒ Object

flush services to the trace-agent, handles services only



73
74
75
76
77
78
79
80
81
82
83
84
 
# File 'lib/ddtrace/writer.rb', line 73

def send_services(services, transport)
  return true if services.empty?

  code = transport.send(:services, services)
  if transport.server_error? code # requeue on server error, skip on success or client error
    @worker.enqueue_service services
    return false
  end

  @services_flushed += 1
  true
end

#send_spans(traces, transport) ⇒ Object

flush spans to the trace-agent, handles spans only



56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
 
# File 'lib/ddtrace/writer.rb', line 56

def send_spans(traces, transport)
  return true if traces.empty?

  code = transport.send(:traces, traces)

  if transport.server_error? code # requeue on server error, skip on success or client error
    traces[0..@buff_size].each do |trace|
      @worker.enqueue_trace trace
    end
    return false
  end

  @traces_flushed += traces.length()
  true
end

#start ⇒ Object

spawns two different workers for spans and services; they share the same transport which is thread-safe



36
37
38
39
40
41
42
43
44
45
46
47
 
# File 'lib/ddtrace/writer.rb', line 36

def start
  @trace_handler = ->(items, transport) { send_spans(items, transport) }
  @service_handler = ->(items, transport) { send_services(items, transport) }
  @worker = Datadog::Workers::AsyncTransport.new(@span_interval,
                                                 @service_interval,
                                                 @transport,
                                                 @buff_size,
                                                 @trace_handler,
                                                 @service_handler)

  @worker.start()
end

#stats ⇒ Object

stats returns a dictionary of stats about the writer.



110
111
112
113
114
115
116
 
# File 'lib/ddtrace/writer.rb', line 110

def stats
  {
    traces_flushed: @traces_flushed,
    services_flushed: @services_flushed,
    transport: @transport.stats
  }
end

#stop ⇒ Object

stops both workers for spans and services.



50
51
52
53
 
# File 'lib/ddtrace/writer.rb', line 50

def stop
  @worker.stop()
  @worker = nil
end

#write(trace, services) ⇒ Object

enqueue the trace for submission to the API



87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
 
# File 'lib/ddtrace/writer.rb', line 87

def write(trace, services)
  # In multiprocess environments, the main process initializes the +Writer+ instance and if
  # the process forks (i.e. a web server like Unicorn or Puma with multiple workers) the new
  # processes will share the same +Writer+ until the first write (COW). Because of that,
  # each process owns a different copy of the +@buffer+ after each write and so the
  # +AsyncTransport+ will not send data to the trace agent.
  #
  # This check ensures that if a process doesn't own the current +Writer+, async workers
  # will be initialized again (but only once for each process).
  pid = Process.pid
  @mutex_after_fork.synchronize do
    if pid != @pid
      @pid = pid
      # we should start threads because the worker doesn't own this
      start()
    end
  end

  @worker.enqueue_trace(trace)
  @worker.enqueue_service(services)
end