ruby-kafka-custom 0.7.7.26

Sign up to get free protection for your applications and to get access to all the features.
Files changed (105) hide show
  1. checksums.yaml +7 -0
  2. data/lib/kafka/async_producer.rb +279 -0
  3. data/lib/kafka/broker.rb +205 -0
  4. data/lib/kafka/broker_info.rb +16 -0
  5. data/lib/kafka/broker_pool.rb +41 -0
  6. data/lib/kafka/broker_uri.rb +43 -0
  7. data/lib/kafka/client.rb +754 -0
  8. data/lib/kafka/cluster.rb +455 -0
  9. data/lib/kafka/compression.rb +43 -0
  10. data/lib/kafka/compressor.rb +85 -0
  11. data/lib/kafka/connection.rb +220 -0
  12. data/lib/kafka/connection_builder.rb +33 -0
  13. data/lib/kafka/consumer.rb +592 -0
  14. data/lib/kafka/consumer_group.rb +208 -0
  15. data/lib/kafka/datadog.rb +413 -0
  16. data/lib/kafka/fetch_operation.rb +115 -0
  17. data/lib/kafka/fetched_batch.rb +54 -0
  18. data/lib/kafka/fetched_batch_generator.rb +117 -0
  19. data/lib/kafka/fetched_message.rb +47 -0
  20. data/lib/kafka/fetched_offset_resolver.rb +48 -0
  21. data/lib/kafka/fetcher.rb +221 -0
  22. data/lib/kafka/gzip_codec.rb +30 -0
  23. data/lib/kafka/heartbeat.rb +25 -0
  24. data/lib/kafka/instrumenter.rb +38 -0
  25. data/lib/kafka/lz4_codec.rb +23 -0
  26. data/lib/kafka/message_buffer.rb +87 -0
  27. data/lib/kafka/offset_manager.rb +248 -0
  28. data/lib/kafka/partitioner.rb +35 -0
  29. data/lib/kafka/pause.rb +92 -0
  30. data/lib/kafka/pending_message.rb +29 -0
  31. data/lib/kafka/pending_message_queue.rb +41 -0
  32. data/lib/kafka/produce_operation.rb +205 -0
  33. data/lib/kafka/producer.rb +504 -0
  34. data/lib/kafka/protocol.rb +217 -0
  35. data/lib/kafka/protocol/add_partitions_to_txn_request.rb +34 -0
  36. data/lib/kafka/protocol/add_partitions_to_txn_response.rb +47 -0
  37. data/lib/kafka/protocol/alter_configs_request.rb +44 -0
  38. data/lib/kafka/protocol/alter_configs_response.rb +49 -0
  39. data/lib/kafka/protocol/api_versions_request.rb +21 -0
  40. data/lib/kafka/protocol/api_versions_response.rb +53 -0
  41. data/lib/kafka/protocol/consumer_group_protocol.rb +19 -0
  42. data/lib/kafka/protocol/create_partitions_request.rb +42 -0
  43. data/lib/kafka/protocol/create_partitions_response.rb +28 -0
  44. data/lib/kafka/protocol/create_topics_request.rb +45 -0
  45. data/lib/kafka/protocol/create_topics_response.rb +26 -0
  46. data/lib/kafka/protocol/decoder.rb +175 -0
  47. data/lib/kafka/protocol/delete_topics_request.rb +33 -0
  48. data/lib/kafka/protocol/delete_topics_response.rb +26 -0
  49. data/lib/kafka/protocol/describe_configs_request.rb +35 -0
  50. data/lib/kafka/protocol/describe_configs_response.rb +73 -0
  51. data/lib/kafka/protocol/describe_groups_request.rb +27 -0
  52. data/lib/kafka/protocol/describe_groups_response.rb +73 -0
  53. data/lib/kafka/protocol/encoder.rb +184 -0
  54. data/lib/kafka/protocol/end_txn_request.rb +29 -0
  55. data/lib/kafka/protocol/end_txn_response.rb +19 -0
  56. data/lib/kafka/protocol/fetch_request.rb +70 -0
  57. data/lib/kafka/protocol/fetch_response.rb +136 -0
  58. data/lib/kafka/protocol/find_coordinator_request.rb +29 -0
  59. data/lib/kafka/protocol/find_coordinator_response.rb +29 -0
  60. data/lib/kafka/protocol/heartbeat_request.rb +27 -0
  61. data/lib/kafka/protocol/heartbeat_response.rb +17 -0
  62. data/lib/kafka/protocol/init_producer_id_request.rb +26 -0
  63. data/lib/kafka/protocol/init_producer_id_response.rb +27 -0
  64. data/lib/kafka/protocol/join_group_request.rb +41 -0
  65. data/lib/kafka/protocol/join_group_response.rb +33 -0
  66. data/lib/kafka/protocol/leave_group_request.rb +25 -0
  67. data/lib/kafka/protocol/leave_group_response.rb +17 -0
  68. data/lib/kafka/protocol/list_groups_request.rb +23 -0
  69. data/lib/kafka/protocol/list_groups_response.rb +35 -0
  70. data/lib/kafka/protocol/list_offset_request.rb +53 -0
  71. data/lib/kafka/protocol/list_offset_response.rb +89 -0
  72. data/lib/kafka/protocol/member_assignment.rb +42 -0
  73. data/lib/kafka/protocol/message.rb +172 -0
  74. data/lib/kafka/protocol/message_set.rb +55 -0
  75. data/lib/kafka/protocol/metadata_request.rb +31 -0
  76. data/lib/kafka/protocol/metadata_response.rb +185 -0
  77. data/lib/kafka/protocol/offset_commit_request.rb +47 -0
  78. data/lib/kafka/protocol/offset_commit_response.rb +29 -0
  79. data/lib/kafka/protocol/offset_fetch_request.rb +36 -0
  80. data/lib/kafka/protocol/offset_fetch_response.rb +56 -0
  81. data/lib/kafka/protocol/produce_request.rb +92 -0
  82. data/lib/kafka/protocol/produce_response.rb +63 -0
  83. data/lib/kafka/protocol/record.rb +88 -0
  84. data/lib/kafka/protocol/record_batch.rb +222 -0
  85. data/lib/kafka/protocol/request_message.rb +26 -0
  86. data/lib/kafka/protocol/sasl_handshake_request.rb +33 -0
  87. data/lib/kafka/protocol/sasl_handshake_response.rb +28 -0
  88. data/lib/kafka/protocol/sync_group_request.rb +33 -0
  89. data/lib/kafka/protocol/sync_group_response.rb +23 -0
  90. data/lib/kafka/round_robin_assignment_strategy.rb +54 -0
  91. data/lib/kafka/sasl/gssapi.rb +76 -0
  92. data/lib/kafka/sasl/oauth.rb +64 -0
  93. data/lib/kafka/sasl/plain.rb +39 -0
  94. data/lib/kafka/sasl/scram.rb +177 -0
  95. data/lib/kafka/sasl_authenticator.rb +61 -0
  96. data/lib/kafka/snappy_codec.rb +25 -0
  97. data/lib/kafka/socket_with_timeout.rb +96 -0
  98. data/lib/kafka/ssl_context.rb +66 -0
  99. data/lib/kafka/ssl_socket_with_timeout.rb +187 -0
  100. data/lib/kafka/statsd.rb +296 -0
  101. data/lib/kafka/tagged_logger.rb +72 -0
  102. data/lib/kafka/transaction_manager.rb +261 -0
  103. data/lib/kafka/transaction_state_machine.rb +72 -0
  104. data/lib/kafka/version.rb +5 -0
  105. metadata +461 -0
@@ -0,0 +1,61 @@
1
+ # frozen_string_literal: true
2
+
3
+ require 'kafka/sasl/plain'
4
+ require 'kafka/sasl/gssapi'
5
+ require 'kafka/sasl/scram'
6
+ require 'kafka/sasl/oauth'
7
+
8
+ module Kafka
9
+ class SaslAuthenticator
10
+ def initialize(logger:, sasl_gssapi_principal:, sasl_gssapi_keytab:,
11
+ sasl_plain_authzid:, sasl_plain_username:, sasl_plain_password:,
12
+ sasl_scram_username:, sasl_scram_password:, sasl_scram_mechanism:,
13
+ sasl_oauth_token_provider:)
14
+ @logger = TaggedLogger.new(logger)
15
+
16
+ @plain = Sasl::Plain.new(
17
+ authzid: sasl_plain_authzid,
18
+ username: sasl_plain_username,
19
+ password: sasl_plain_password,
20
+ logger: @logger,
21
+ )
22
+
23
+ @gssapi = Sasl::Gssapi.new(
24
+ principal: sasl_gssapi_principal,
25
+ keytab: sasl_gssapi_keytab,
26
+ logger: @logger,
27
+ )
28
+
29
+ @scram = Sasl::Scram.new(
30
+ username: sasl_scram_username,
31
+ password: sasl_scram_password,
32
+ mechanism: sasl_scram_mechanism,
33
+ logger: @logger,
34
+ )
35
+
36
+ @oauth = Sasl::OAuth.new(
37
+ token_provider: sasl_oauth_token_provider,
38
+ logger: @logger,
39
+ )
40
+
41
+ @mechanism = [@gssapi, @plain, @scram, @oauth].find(&:configured?)
42
+ end
43
+
44
+ def enabled?
45
+ !@mechanism.nil?
46
+ end
47
+
48
+ def authenticate!(connection)
49
+ return unless enabled?
50
+
51
+ ident = @mechanism.ident
52
+ response = connection.send_request(Kafka::Protocol::SaslHandshakeRequest.new(ident))
53
+
54
+ unless response.error_code == 0 && response.enabled_mechanisms.include?(ident)
55
+ raise Kafka::Error, "#{ident} is not supported."
56
+ end
57
+
58
+ @mechanism.authenticate!(connection.to_s, connection.encoder, connection.decoder)
59
+ end
60
+ end
61
+ end
@@ -0,0 +1,25 @@
1
+ # frozen_string_literal: true
2
+
3
+ module Kafka
4
+ class SnappyCodec
5
+ def codec_id
6
+ 2
7
+ end
8
+
9
+ def load
10
+ require "snappy"
11
+ rescue LoadError
12
+ raise LoadError,
13
+ "Using snappy compression requires adding a dependency on the `snappy` gem to your Gemfile."
14
+ end
15
+
16
+ def compress(data)
17
+ Snappy.deflate(data)
18
+ end
19
+
20
+ def decompress(data)
21
+ buffer = StringIO.new(data)
22
+ Snappy::Reader.new(buffer).read
23
+ end
24
+ end
25
+ end
@@ -0,0 +1,96 @@
1
+ # frozen_string_literal: true
2
+
3
+ require "socket"
4
+
5
+ module Kafka
6
+
7
+ # Opens sockets in a non-blocking fashion, ensuring that we're not stalling
8
+ # for long periods of time.
9
+ #
10
+ # It's possible to set timeouts for connecting to the server, for reading data,
11
+ # and for writing data. Whenever a timeout is exceeded, Errno::ETIMEDOUT is
12
+ # raised.
13
+ #
14
+ class SocketWithTimeout
15
+
16
+ # Opens a socket.
17
+ #
18
+ # @param host [String]
19
+ # @param port [Integer]
20
+ # @param connect_timeout [Integer] the connection timeout, in seconds.
21
+ # @param timeout [Integer] the read and write timeout, in seconds.
22
+ # @raise [Errno::ETIMEDOUT] if the timeout is exceeded.
23
+ def initialize(host, port, connect_timeout: nil, timeout: nil)
24
+ addr = Socket.getaddrinfo(host, nil)
25
+ sockaddr = Socket.pack_sockaddr_in(port, addr[0][3])
26
+
27
+ @timeout = timeout
28
+
29
+ @socket = Socket.new(Socket.const_get(addr[0][0]), Socket::SOCK_STREAM, 0)
30
+ @socket.setsockopt(Socket::IPPROTO_TCP, Socket::TCP_NODELAY, 1)
31
+
32
+ begin
33
+ # Initiate the socket connection in the background. If it doesn't fail
34
+ # immediately it will raise an IO::WaitWritable (Errno::EINPROGRESS)
35
+ # indicating the connection is in progress.
36
+ @socket.connect_nonblock(sockaddr)
37
+ rescue IO::WaitWritable
38
+ # IO.select will block until the socket is writable or the timeout
39
+ # is exceeded, whichever comes first.
40
+ unless IO.select(nil, [@socket], nil, connect_timeout)
41
+ # IO.select returns nil when the socket is not ready before timeout
42
+ # seconds have elapsed
43
+ @socket.close
44
+ raise Errno::ETIMEDOUT
45
+ end
46
+
47
+ begin
48
+ # Verify there is now a good connection.
49
+ @socket.connect_nonblock(sockaddr)
50
+ rescue Errno::EISCONN
51
+ # The socket is connected, we're good!
52
+ end
53
+ end
54
+ end
55
+
56
+ # Reads bytes from the socket, possible with a timeout.
57
+ #
58
+ # @param num_bytes [Integer] the number of bytes to read.
59
+ # @raise [Errno::ETIMEDOUT] if the timeout is exceeded.
60
+ # @return [String] the data that was read from the socket.
61
+ def read(num_bytes)
62
+ unless IO.select([@socket], nil, nil, @timeout)
63
+ raise Errno::ETIMEDOUT
64
+ end
65
+
66
+ @socket.read(num_bytes)
67
+ rescue IO::EAGAINWaitReadable
68
+ retry
69
+ end
70
+
71
+ # Writes bytes to the socket, possible with a timeout.
72
+ #
73
+ # @param bytes [String] the data that should be written to the socket.
74
+ # @raise [Errno::ETIMEDOUT] if the timeout is exceeded.
75
+ # @return [Integer] the number of bytes written.
76
+ def write(bytes)
77
+ unless IO.select(nil, [@socket], nil, @timeout)
78
+ raise Errno::ETIMEDOUT
79
+ end
80
+
81
+ @socket.write(bytes)
82
+ end
83
+
84
+ def close
85
+ @socket.close
86
+ end
87
+
88
+ def closed?
89
+ @socket.closed?
90
+ end
91
+
92
+ def set_encoding(encoding)
93
+ @socket.set_encoding(encoding)
94
+ end
95
+ end
96
+ end
@@ -0,0 +1,66 @@
1
+ # frozen_string_literal: true
2
+
3
+ require "openssl"
4
+
5
+ module Kafka
6
+ module SslContext
7
+ CLIENT_CERT_DELIMITER = "\n-----END CERTIFICATE-----\n"
8
+
9
+ def self.build(ca_cert_file_path: nil, ca_cert: nil, client_cert: nil, client_cert_key: nil, client_cert_key_password: nil, client_cert_chain: nil, ca_certs_from_system: nil)
10
+ return nil unless ca_cert_file_path || ca_cert || client_cert || client_cert_key || client_cert_key_password || client_cert_chain || ca_certs_from_system
11
+
12
+ ssl_context = OpenSSL::SSL::SSLContext.new
13
+
14
+ if client_cert && client_cert_key
15
+ if client_cert_key_password
16
+ cert_key = OpenSSL::PKey.read(client_cert_key, client_cert_key_password)
17
+ else
18
+ cert_key = OpenSSL::PKey.read(client_cert_key)
19
+ end
20
+ context_params = {
21
+ cert: OpenSSL::X509::Certificate.new(client_cert),
22
+ key: cert_key
23
+ }
24
+ if client_cert_chain
25
+ certs = []
26
+ client_cert_chain.split(CLIENT_CERT_DELIMITER).each do |cert|
27
+ cert += CLIENT_CERT_DELIMITER
28
+ certs << OpenSSL::X509::Certificate.new(cert)
29
+ end
30
+ context_params[:extra_chain_cert] = certs
31
+ end
32
+ ssl_context.set_params(context_params)
33
+ elsif client_cert && !client_cert_key
34
+ raise ArgumentError, "Kafka client initialized with `ssl_client_cert` but no `ssl_client_cert_key`. Please provide both."
35
+ elsif !client_cert && client_cert_key
36
+ raise ArgumentError, "Kafka client initialized with `ssl_client_cert_key`, but no `ssl_client_cert`. Please provide both."
37
+ elsif client_cert_chain && !client_cert
38
+ raise ArgumentError, "Kafka client initialized with `ssl_client_cert_chain`, but no `ssl_client_cert`. Please provide cert, key and chain."
39
+ elsif client_cert_chain && !client_cert_key
40
+ raise ArgumentError, "Kafka client initialized with `ssl_client_cert_chain`, but no `ssl_client_cert_key`. Please provide cert, key and chain."
41
+ elsif client_cert_key_password && !client_cert_key
42
+ raise ArgumentError, "Kafka client initialized with `ssl_client_cert_key_password`, but no `ssl_client_cert_key`. Please provide both."
43
+ end
44
+
45
+ puts "----------------------------------"
46
+ puts OpenSSL::OPENSSL_LIBRARY_VERSION
47
+ puts "----------------------------------"
48
+
49
+ if ca_cert || ca_cert_file_path || ca_certs_from_system
50
+ store = OpenSSL::X509::Store.new
51
+ Array(ca_cert).each do |cert|
52
+ store.add_cert(OpenSSL::X509::Certificate.new(cert))
53
+ end
54
+ if ca_cert_file_path
55
+ store.add_file(ca_cert_file_path)
56
+ end
57
+ if ca_certs_from_system
58
+ store.set_default_paths
59
+ end
60
+ ssl_context.cert_store = store
61
+ end
62
+
63
+ ssl_context
64
+ end
65
+ end
66
+ end
@@ -0,0 +1,187 @@
1
+ # frozen_string_literal: true
2
+
3
+ require "socket"
4
+
5
+ module Kafka
6
+
7
+ # Opens sockets in a non-blocking fashion, ensuring that we're not stalling
8
+ # for long periods of time.
9
+ #
10
+ # It's possible to set timeouts for connecting to the server, for reading data,
11
+ # and for writing data. Whenever a timeout is exceeded, Errno::ETIMEDOUT is
12
+ # raised.
13
+ #
14
+ class SSLSocketWithTimeout
15
+
16
+ # Opens a socket.
17
+ #
18
+ # @param host [String]
19
+ # @param port [Integer]
20
+ # @param connect_timeout [Integer] the connection timeout, in seconds.
21
+ # @param timeout [Integer] the read and write timeout, in seconds.
22
+ # @param ssl_context [OpenSSL::SSL::SSLContext] which SSLContext the ssl connection should use
23
+ # @raise [Errno::ETIMEDOUT] if the timeout is exceeded.
24
+ def initialize(host, port, connect_timeout: nil, timeout: nil, ssl_context:)
25
+ addr = Socket.getaddrinfo(host, nil)
26
+ sockaddr = Socket.pack_sockaddr_in(port, addr[0][3])
27
+
28
+ @connect_timeout = connect_timeout
29
+ @timeout = timeout
30
+
31
+ @tcp_socket = Socket.new(Socket.const_get(addr[0][0]), Socket::SOCK_STREAM, 0)
32
+ @tcp_socket.setsockopt(Socket::IPPROTO_TCP, Socket::TCP_NODELAY, 1)
33
+
34
+ # first initiate the TCP socket
35
+ begin
36
+ # Initiate the socket connection in the background. If it doesn't fail
37
+ # immediately it will raise an IO::WaitWritable (Errno::EINPROGRESS)
38
+ # indicating the connection is in progress.
39
+ @tcp_socket.connect_nonblock(sockaddr)
40
+ rescue IO::WaitWritable
41
+ # select will block until the socket is writable or the timeout
42
+ # is exceeded, whichever comes first.
43
+ unless select_with_timeout(@tcp_socket, :connect_write)
44
+ # select returns nil when the socket is not ready before timeout
45
+ # seconds have elapsed
46
+ @tcp_socket.close
47
+ raise Errno::ETIMEDOUT
48
+ end
49
+
50
+ begin
51
+ # Verify there is now a good connection.
52
+ @tcp_socket.connect_nonblock(sockaddr)
53
+ rescue Errno::EISCONN
54
+ # The socket is connected, we're good!
55
+ end
56
+ end
57
+
58
+ # once that's connected, we can start initiating the ssl socket
59
+ @ssl_socket = OpenSSL::SSL::SSLSocket.new(@tcp_socket, ssl_context)
60
+
61
+ begin
62
+ # Initiate the socket connection in the background. If it doesn't fail
63
+ # immediately it will raise an IO::WaitWritable (Errno::EINPROGRESS)
64
+ # indicating the connection is in progress.
65
+ # Unlike waiting for a tcp socket to connect, you can't time out ssl socket
66
+ # connections during the connect phase properly, because IO.select only partially works.
67
+ # Instead, you have to retry.
68
+ @ssl_socket.connect_nonblock
69
+ rescue Errno::EAGAIN, Errno::EWOULDBLOCK, IO::WaitReadable
70
+ if select_with_timeout(@ssl_socket, :connect_read)
71
+ retry
72
+ else
73
+ @ssl_socket.close
74
+ close
75
+ raise Errno::ETIMEDOUT
76
+ end
77
+ rescue IO::WaitWritable
78
+ if select_with_timeout(@ssl_socket, :connect_write)
79
+ retry
80
+ else
81
+ close
82
+ raise Errno::ETIMEDOUT
83
+ end
84
+ end
85
+ end
86
+
87
+ # Reads bytes from the socket, possible with a timeout.
88
+ #
89
+ # @param num_bytes [Integer] the number of bytes to read.
90
+ # @raise [Errno::ETIMEDOUT] if the timeout is exceeded.
91
+ # @return [String] the data that was read from the socket.
92
+ def read(num_bytes)
93
+ buffer = String.new
94
+
95
+ until buffer.length >= num_bytes
96
+ begin
97
+ # Unlike plain TCP sockets, SSL sockets don't support IO.select
98
+ # properly.
99
+ # Instead, timeouts happen on a per read basis, and we have to
100
+ # catch exceptions from read_nonblock and gradually build up
101
+ # our read buffer.
102
+ buffer << @ssl_socket.read_nonblock(num_bytes - buffer.length)
103
+ rescue IO::WaitReadable
104
+ if select_with_timeout(@ssl_socket, :read)
105
+ retry
106
+ else
107
+ raise Errno::ETIMEDOUT
108
+ end
109
+ rescue IO::WaitWritable
110
+ if select_with_timeout(@ssl_socket, :write)
111
+ retry
112
+ else
113
+ raise Errno::ETIMEDOUT
114
+ end
115
+ end
116
+ end
117
+
118
+ buffer
119
+ end
120
+
121
+ # Writes bytes to the socket, possible with a timeout.
122
+ #
123
+ # @param bytes [String] the data that should be written to the socket.
124
+ # @raise [Errno::ETIMEDOUT] if the timeout is exceeded.
125
+ # @return [Integer] the number of bytes written.
126
+ def write(bytes)
127
+ loop do
128
+ written = 0
129
+ begin
130
+ # unlike plain tcp sockets, ssl sockets don't support IO.select
131
+ # properly.
132
+ # Instead, timeouts happen on a per write basis, and we have to
133
+ # catch exceptions from write_nonblock, and gradually build up
134
+ # our write buffer.
135
+ written += @ssl_socket.write_nonblock(bytes)
136
+ rescue Errno::EFAULT => error
137
+ raise error
138
+ rescue OpenSSL::SSL::SSLError, Errno::EAGAIN, Errno::EWOULDBLOCK, IO::WaitWritable => error
139
+ if error.is_a?(OpenSSL::SSL::SSLError) && error.message == 'write would block'
140
+ if select_with_timeout(@ssl_socket, :write)
141
+ retry
142
+ else
143
+ raise Errno::ETIMEDOUT
144
+ end
145
+ else
146
+ raise error
147
+ end
148
+ end
149
+
150
+ # Fast, common case.
151
+ break if written == bytes.size
152
+
153
+ # This takes advantage of the fact that most ruby implementations
154
+ # have Copy-On-Write strings. Thusly why requesting a subrange
155
+ # of data, we actually don't copy data because the new string
156
+ # simply references a subrange of the original.
157
+ bytes = bytes[written, bytes.size]
158
+ end
159
+ end
160
+
161
+ def close
162
+ @tcp_socket.close
163
+ @ssl_socket.close
164
+ end
165
+
166
+ def closed?
167
+ @tcp_socket.closed? || @ssl_socket.closed?
168
+ end
169
+
170
+ def set_encoding(encoding)
171
+ @tcp_socket.set_encoding(encoding)
172
+ end
173
+
174
+ def select_with_timeout(socket, type)
175
+ case type
176
+ when :connect_read
177
+ IO.select([socket], nil, nil, @connect_timeout)
178
+ when :connect_write
179
+ IO.select(nil, [socket], nil, @connect_timeout)
180
+ when :read
181
+ IO.select([socket], nil, nil, @timeout)
182
+ when :write
183
+ IO.select(nil, [socket], nil, @timeout)
184
+ end
185
+ end
186
+ end
187
+ end
@@ -0,0 +1,296 @@
1
+ # frozen_string_literal: true
2
+
3
+ begin
4
+ require "statsd"
5
+ rescue LoadError
6
+ $stderr.puts "In order to report Kafka client metrics to Statsd you need to install the `statsd-ruby` gem."
7
+ raise
8
+ end
9
+
10
+ require "active_support/subscriber"
11
+
12
+ module Kafka
13
+ # Reports operational metrics to a Statsd agent.
14
+ #
15
+ # require "kafka/statsd"
16
+ #
17
+ # # Default is "ruby_kafka".
18
+ # Kafka::Statsd.namespace = "custom-namespace"
19
+ #
20
+ # # Default is "127.0.0.1".
21
+ # Kafka::Statsd.host = "statsd.something.com"
22
+ #
23
+ # # Default is 8125.
24
+ # Kafka::Statsd.port = 1234
25
+ #
26
+ # Once the file has been required, no further configuration is needed – all operational
27
+ # metrics are automatically emitted.
28
+ module Statsd
29
+ DEFAULT_NAMESPACE = "ruby_kafka"
30
+ DEFAULT_HOST = '127.0.0.1'
31
+ DEFAULT_PORT = 8125
32
+
33
+ def self.statsd
34
+ @statsd ||= ::Statsd.new(DEFAULT_HOST, DEFAULT_PORT).tap { |sd| sd.namespace = DEFAULT_NAMESPACE }
35
+ end
36
+
37
+ def self.host=(host)
38
+ statsd.host = host
39
+ statsd.connect if statsd.respond_to?(:connect)
40
+ end
41
+
42
+ def self.port=(port)
43
+ statsd.port = port
44
+ statsd.connect if statsd.respond_to?(:connect)
45
+ end
46
+
47
+ def self.namespace=(namespace)
48
+ statsd.namespace = namespace
49
+ end
50
+
51
+ class StatsdSubscriber < ActiveSupport::Subscriber
52
+ private
53
+
54
+ %w[increment count timing gauge].each do |type|
55
+ define_method(type) do |*args|
56
+ Kafka::Statsd.statsd.send(type, *args)
57
+ end
58
+ end
59
+ end
60
+
61
+ class ConnectionSubscriber < StatsdSubscriber
62
+ def request(event)
63
+ client = event.payload.fetch(:client_id)
64
+ api = event.payload.fetch(:api, "unknown")
65
+ request_size = event.payload.fetch(:request_size, 0)
66
+ response_size = event.payload.fetch(:response_size, 0)
67
+ broker = event.payload.fetch(:broker_host)
68
+
69
+ timing("api.#{client}.#{api}.#{broker}.latency", event.duration)
70
+ increment("api.#{client}.#{api}.#{broker}.calls")
71
+
72
+ timing("api.#{client}.#{api}.#{broker}.request_size", request_size)
73
+ timing("api.#{client}.#{api}.#{broker}.response_size", response_size)
74
+
75
+ if event.payload.key?(:exception)
76
+ increment("api.#{client}.#{api}.#{broker}.errors")
77
+ end
78
+ end
79
+
80
+ attach_to "connection.kafka"
81
+ end
82
+
83
+ class ConsumerSubscriber < StatsdSubscriber
84
+ def process_message(event)
85
+ offset_lag = event.payload.fetch(:offset_lag)
86
+ create_time = event.payload.fetch(:create_time)
87
+ client = event.payload.fetch(:client_id)
88
+ group_id = event.payload.fetch(:group_id)
89
+ topic = event.payload.fetch(:topic)
90
+ partition = event.payload.fetch(:partition)
91
+
92
+ time_lag = create_time && ((Time.now - create_time) * 1000).to_i
93
+
94
+ if event.payload.key?(:exception)
95
+ increment("consumer.#{client}.#{group_id}.#{topic}.#{partition}.process_message.errors")
96
+ else
97
+ timing("consumer.#{client}.#{group_id}.#{topic}.#{partition}.process_message.latency", event.duration)
98
+ increment("consumer.#{client}.#{group_id}.#{topic}.#{partition}.messages")
99
+ end
100
+
101
+ gauge("consumer.#{client}.#{group_id}.#{topic}.#{partition}.lag", offset_lag)
102
+
103
+ # Not all messages have timestamps.
104
+ if time_lag
105
+ gauge("consumer.#{client}.#{group_id}.#{topic}.#{partition}.time_lag", time_lag)
106
+ end
107
+ end
108
+
109
+ def process_batch(event)
110
+ messages = event.payload.fetch(:message_count)
111
+ client = event.payload.fetch(:client_id)
112
+ group_id = event.payload.fetch(:group_id)
113
+ topic = event.payload.fetch(:topic)
114
+ partition = event.payload.fetch(:partition)
115
+
116
+ if event.payload.key?(:exception)
117
+ increment("consumer.#{client}.#{group_id}.#{topic}.#{partition}.process_batch.errors")
118
+ else
119
+ timing("consumer.#{client}.#{group_id}.#{topic}.#{partition}.process_batch.latency", event.duration)
120
+ count("consumer.#{client}.#{group_id}.#{topic}.#{partition}.messages", messages)
121
+ end
122
+ end
123
+
124
+ def fetch_batch(event)
125
+ lag = event.payload.fetch(:offset_lag)
126
+ batch_size = event.payload.fetch(:message_count)
127
+ client = event.payload.fetch(:client_id)
128
+ group_id = event.payload.fetch(:group_id)
129
+ topic = event.payload.fetch(:topic)
130
+ partition = event.payload.fetch(:partition)
131
+
132
+ count("consumer.#{client}.#{group_id}.#{topic}.#{partition}.batch_size", batch_size)
133
+ gauge("consumer.#{client}.#{group_id}.#{topic}.#{partition}.lag", lag)
134
+ end
135
+
136
+ def join_group(event)
137
+ client = event.payload.fetch(:client_id)
138
+ group_id = event.payload.fetch(:group_id)
139
+
140
+ timing("consumer.#{client}.#{group_id}.join_group", event.duration)
141
+
142
+ if event.payload.key?(:exception)
143
+ increment("consumer.#{client}.#{group_id}.join_group.errors")
144
+ end
145
+ end
146
+
147
+ def sync_group(event)
148
+ client = event.payload.fetch(:client_id)
149
+ group_id = event.payload.fetch(:group_id)
150
+
151
+ timing("consumer.#{client}.#{group_id}.sync_group", event.duration)
152
+
153
+ if event.payload.key?(:exception)
154
+ increment("consumer.#{client}.#{group_id}.sync_group.errors")
155
+ end
156
+ end
157
+
158
+ def leave_group(event)
159
+ client = event.payload.fetch(:client_id)
160
+ group_id = event.payload.fetch(:group_id)
161
+
162
+ timing("consumer.#{client}.#{group_id}.leave_group", event.duration)
163
+
164
+ if event.payload.key?(:exception)
165
+ increment("consumer.#{client}.#{group_id}.leave_group.errors")
166
+ end
167
+ end
168
+
169
+ def pause_status(event)
170
+ client = event.payload.fetch(:client_id)
171
+ group_id = event.payload.fetch(:group_id)
172
+ topic = event.payload.fetch(:topic)
173
+ partition = event.payload.fetch(:partition)
174
+
175
+ duration = event.payload.fetch(:duration)
176
+
177
+ gauge("consumer.#{client}.#{group_id}.#{topic}.#{partition}.pause.duration", duration)
178
+ end
179
+
180
+ attach_to "consumer.kafka"
181
+ end
182
+
183
+ class ProducerSubscriber < StatsdSubscriber
184
+ def produce_message(event)
185
+ client = event.payload.fetch(:client_id)
186
+ topic = event.payload.fetch(:topic)
187
+ message_size = event.payload.fetch(:message_size)
188
+ buffer_size = event.payload.fetch(:buffer_size)
189
+ max_buffer_size = event.payload.fetch(:max_buffer_size)
190
+ buffer_fill_ratio = buffer_size.to_f / max_buffer_size.to_f
191
+ buffer_fill_percentage = buffer_fill_ratio * 100.0
192
+
193
+ # This gets us the write rate.
194
+ increment("producer.#{client}.#{topic}.produce.messages")
195
+
196
+ timing("producer.#{client}.#{topic}.produce.message_size", message_size)
197
+
198
+ # This gets us the avg/max buffer size per producer.
199
+ timing("producer.#{client}.buffer.size", buffer_size)
200
+
201
+ # This gets us the avg/max buffer fill ratio per producer.
202
+ timing("producer.#{client}.buffer.fill_ratio", buffer_fill_ratio)
203
+ timing("producer.#{client}.buffer.fill_percentage", buffer_fill_percentage)
204
+ end
205
+
206
+ def buffer_overflow(event)
207
+ client = event.payload.fetch(:client_id)
208
+ topic = event.payload.fetch(:topic)
209
+
210
+ increment("producer.#{client}.#{topic}.produce.errors")
211
+ end
212
+
213
+ def deliver_messages(event)
214
+ client = event.payload.fetch(:client_id)
215
+ message_count = event.payload.fetch(:delivered_message_count)
216
+ attempts = event.payload.fetch(:attempts)
217
+
218
+ if event.payload.key?(:exception)
219
+ increment("producer.#{client}.deliver.errors")
220
+ end
221
+
222
+ timing("producer.#{client}.deliver.latency", event.duration)
223
+
224
+ # Messages delivered to Kafka:
225
+ count("producer.#{client}.deliver.messages", message_count)
226
+
227
+ # Number of attempts to deliver messages:
228
+ timing("producer.#{client}.deliver.attempts", attempts)
229
+ end
230
+
231
+ def ack_message(event)
232
+ client = event.payload.fetch(:client_id)
233
+ topic = event.payload.fetch(:topic)
234
+
235
+ # Number of messages ACK'd for the topic.
236
+ increment("producer.#{client}.#{topic}.ack.messages")
237
+
238
+ # Histogram of delay between a message being produced and it being ACK'd.
239
+ timing("producer.#{client}.#{topic}.ack.delay", event.payload.fetch(:delay))
240
+ end
241
+
242
+ def topic_error(event)
243
+ client = event.payload.fetch(:client_id)
244
+ topic = event.payload.fetch(:topic)
245
+
246
+ increment("producer.#{client}.#{topic}.ack.errors")
247
+ end
248
+
249
+ attach_to "producer.kafka"
250
+ end
251
+
252
+ class AsyncProducerSubscriber < StatsdSubscriber
253
+ def enqueue_message(event)
254
+ client = event.payload.fetch(:client_id)
255
+ topic = event.payload.fetch(:topic)
256
+ queue_size = event.payload.fetch(:queue_size)
257
+ max_queue_size = event.payload.fetch(:max_queue_size)
258
+ queue_fill_ratio = queue_size.to_f / max_queue_size.to_f
259
+
260
+ # This gets us the avg/max queue size per producer.
261
+ timing("async_producer.#{client}.#{topic}.queue.size", queue_size)
262
+
263
+ # This gets us the avg/max queue fill ratio per producer.
264
+ timing("async_producer.#{client}.#{topic}.queue.fill_ratio", queue_fill_ratio)
265
+ end
266
+
267
+ def buffer_overflow(event)
268
+ client = event.payload.fetch(:client_id)
269
+ topic = event.payload.fetch(:topic)
270
+
271
+ increment("async_producer.#{client}.#{topic}.produce.errors")
272
+ end
273
+
274
+ def drop_messages(event)
275
+ client = event.payload.fetch(:client_id)
276
+ message_count = event.payload.fetch(:message_count)
277
+
278
+ count("async_producer.#{client}.dropped_messages", message_count)
279
+ end
280
+
281
+ attach_to "async_producer.kafka"
282
+ end
283
+
284
+ class FetcherSubscriber < StatsdSubscriber
285
+ def loop(event)
286
+ queue_size = event.payload.fetch(:queue_size)
287
+ client = event.payload.fetch(:client_id)
288
+ group_id = event.payload.fetch(:group_id)
289
+
290
+ gauge("fetcher.#{client}.#{group_id}.queue_size", queue_size)
291
+ end
292
+
293
+ attach_to "fetcher.kafka"
294
+ end
295
+ end
296
+ end