Class: Kafka::Cluster

Inherits:
Object
  • Object
show all
Defined in:
lib/kafka/cluster.rb

Overview

A cluster represents the state of a Kafka cluster. It needs to be initialized with a non-empty list of seed brokers. The first seed broker that the cluster can connect to will be asked for the cluster metadata, which allows the cluster to map topic partitions to the current leader for those partitions.

Instance Method Summary collapse

Constructor Details

#initialize(seed_brokers:, broker_pool:, logger:) ⇒ Cluster

Initializes a Cluster with a set of seed brokers.

The cluster will try to fetch cluster metadata from one of the brokers.

Parameters:



21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
 
# File 'lib/kafka/cluster.rb', line 21

def initialize(seed_brokers:, broker_pool:, logger:)
  if seed_brokers.empty?
    raise ArgumentError, "At least one seed broker must be configured"
  end

  @logger = logger
  @seed_brokers = seed_brokers
  @broker_pool = broker_pool
  @cluster_info = nil
  @stale = true

  # This is the set of topics we need metadata for. If empty, metadata for
  # all topics will be fetched.
  @target_topics = Set.new
end

Instance Method Details

#add_target_topics(topics) ⇒ nil

Adds a list of topics to the target list. Only the topics on this list will be queried for metadata.

Parameters:

  • topics (Array<String>)

Returns:

  • (nil) 


42
43
44
45
46
47
48
49
50
51
52
 
# File 'lib/kafka/cluster.rb', line 42

def add_target_topics(topics)
  new_topics = Set.new(topics) - @target_topics

  unless new_topics.empty?
    @logger.info "New topics added to target list: #{new_topics.to_a.join(', ')}"

    @target_topics.merge(new_topics)

    refresh_metadata!
  end
end

#alter_topic(name, configs = {}) ⇒ Object 



240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
 
# File 'lib/kafka/cluster.rb', line 240

def alter_topic(name, configs = {})
  options = {
    resources: [[Kafka::Protocol::RESOURCE_TYPE_TOPIC, name, configs]]
  }

  broker = controller_broker

  @logger.info "Altering the config for topic `#{name}` using controller broker #{broker}"

  response = broker.alter_configs(**options)

  response.resources.each do |resource|
    Protocol.handle_error(resource.error_code, resource.error_message)
  end

  nil
end

#api_info(api_key) ⇒ Object 



54
55
56
 
# File 'lib/kafka/cluster.rb', line 54

def api_info(api_key)
  apis.find {|api| api.api_key == api_key }
end

#apis ⇒ Object 



69
70
71
72
73
74
75
76
77
78
 
# File 'lib/kafka/cluster.rb', line 69

def apis
  @apis ||=
    begin
      response = random_broker.api_versions

      Protocol.handle_error(response.error_code)

      response.apis
    end
end

#clear_target_topics ⇒ nil

Clears the list of target topics.

Returns:

  • (nil)

See Also:



84
85
86
87
 
# File 'lib/kafka/cluster.rb', line 84

def clear_target_topics
  @target_topics.clear
  refresh_metadata!
end

#create_partitions_for(name, num_partitions:, timeout:) ⇒ Object 



265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
 
# File 'lib/kafka/cluster.rb', line 265

def create_partitions_for(name, num_partitions:, timeout:)
  options = {
    topics: [[name, num_partitions, nil]],
    timeout: timeout
  }

  broker = controller_broker

  @logger.info "Creating #{num_partitions} partition(s) for topic `#{name}` using controller broker #{broker}"

  response = broker.create_partitions(**options)

  response.errors.each do |topic, error_code, error_message|
    Protocol.handle_error(error_code, error_message)
  end
  mark_as_stale!

  @logger.info "Topic `#{name}` was updated"
end

#create_topic(name, num_partitions:, replication_factor:, timeout:, config:) ⇒ Object 



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
 
# File 'lib/kafka/cluster.rb', line 161

def create_topic(name, num_partitions:, replication_factor:, timeout:, config:)
  options = {
    topics: {
      name => {
        num_partitions: num_partitions,
        replication_factor: replication_factor,
        config: config,
      }
    },
    timeout: timeout,
  }

  broker = controller_broker

  @logger.info "Creating topic `#{name}` using controller broker #{broker}"

  response = broker.create_topics(**options)

  response.errors.each do |topic, error_code|
    Protocol.handle_error(error_code)
  end

  begin
    partitions_for(name).each do |info|
      Protocol.handle_error(info.partition_error_code)
    end
  rescue Kafka::LeaderNotAvailable
    @logger.warn "Leader not yet available for `#{name}`, waiting 1s..."
    sleep 1

    retry
  rescue Kafka::UnknownTopicOrPartition
    @logger.warn "Topic `#{name}` not yet created, waiting 1s..."
    sleep 1

    retry
  end

  @logger.info "Topic `#{name}` was created"
end

#delete_topic(name, timeout:) ⇒ Object 



202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
 
# File 'lib/kafka/cluster.rb', line 202

def delete_topic(name, timeout:)
  options = {
    topics: [name],
    timeout: timeout,
  }

  broker = controller_broker

  @logger.info "Deleting topic `#{name}` using controller broker #{broker}"

  response = broker.delete_topics(**options)

  response.errors.each do |topic, error_code|
    Protocol.handle_error(error_code)
  end

  @logger.info "Topic `#{name}` was deleted"
end

#describe_group(group_id) ⇒ Object 



258
259
260
261
262
263
 
# File 'lib/kafka/cluster.rb', line 258

def describe_group(group_id)
  response = get_group_coordinator(group_id: group_id).describe_groups(group_ids: [group_id])
  group = response.groups.first
  Protocol.handle_error(group.error_code)
  group
end

#describe_topic(name, configs = []) ⇒ Object 



221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
 
# File 'lib/kafka/cluster.rb', line 221

def describe_topic(name, configs = [])
  options = {
    resources: [[Kafka::Protocol::RESOURCE_TYPE_TOPIC, name, configs]]
  }
  broker = controller_broker

  @logger.info "Fetching topic `#{name}`'s configs using controller broker #{broker}"

  response = broker.describe_configs(**options)

  response.resources.each do |resource|
    Protocol.handle_error(resource.error_code, resource.error_message)
  end
  topic_description = response.resources.first
  topic_description.configs.each_with_object({}) do |config, hash|
    hash[config.name] = config.value
  end
end

#disconnect ⇒ Object 



356
357
358
 
# File 'lib/kafka/cluster.rb', line 356

def disconnect
  @broker_pool.close
end

#get_group_coordinator(group_id:) ⇒ Object

Raises:



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
 
# File 'lib/kafka/cluster.rb', line 111

def get_group_coordinator(group_id:)
  @logger.debug "Getting group coordinator for `#{group_id}`"

  refresh_metadata_if_necessary!

  cluster_info.brokers.each do |broker_info|
    begin
      broker = connect_to_broker(broker_info.node_id)
      response = broker.find_group_coordinator(group_id: group_id)

      Protocol.handle_error(response.error_code)

      coordinator_id = response.coordinator_id

      @logger.debug "Coordinator for group `#{group_id}` is #{coordinator_id}. Connecting..."

      # It's possible that a new broker is introduced to the cluster and
      # becomes the coordinator before we have a chance to refresh_metadata.
      coordinator = begin
        connect_to_broker(coordinator_id)
      rescue Kafka::NoSuchBroker
        @logger.debug "Broker #{coordinator_id} missing from broker cache, refreshing"
        refresh_metadata!
        connect_to_broker(coordinator_id)
      end

      @logger.debug "Connected to coordinator: #{coordinator} for group `#{group_id}`"

      return coordinator
    rescue GroupCoordinatorNotAvailable
      @logger.debug "Coordinator not available; retrying in 1s"
      sleep 1
      retry
    rescue ConnectionError => e
      @logger.error "Failed to get group coordinator info from #{broker}: #{e}"
    end
  end

  raise Kafka::Error, "Failed to find group coordinator"
end

#get_leader(topic, partition) ⇒ Broker

Finds the broker acting as the leader of the given topic and partition.

Parameters:

  • topic (String)
  • partition (Integer)

Returns:

  • (Broker) —

    the broker that's currently leader.



107
108
109
 
# File 'lib/kafka/cluster.rb', line 107

def get_leader(topic, partition)
  connect_to_broker(get_leader_id(topic, partition))
end

#list_groups ⇒ Object 



347
348
349
350
351
352
353
354
 
# File 'lib/kafka/cluster.rb', line 347

def list_groups
  refresh_metadata_if_necessary!
  cluster_info.brokers.map do |broker|
    response = connect_to_broker(broker.node_id).list_groups
    Protocol.handle_error(response.error_code)
    response.groups.map(&:group_id)
  end.flatten.uniq
end

#list_topics ⇒ Object

Lists all topics in the cluster.



340
341
342
343
344
345
 
# File 'lib/kafka/cluster.rb', line 340

def list_topics
  response = random_broker.fetch_metadata(topics: nil)
  response.topics.select do |topic|
    topic.topic_error_code == 0
  end.map(&:topic_name)
end

#mark_as_stale! ⇒ Object 



89
90
91
 
# File 'lib/kafka/cluster.rb', line 89

def mark_as_stale!
  @stale = true
end

#partitions_for(topic) ⇒ Object 



152
153
154
155
156
157
158
159
 
# File 'lib/kafka/cluster.rb', line 152

def partitions_for(topic)
  add_target_topics([topic])
  refresh_metadata_if_necessary!
  cluster_info.partitions_for(topic)
rescue Kafka::ProtocolError
  mark_as_stale!
  raise
end

#refresh_metadata! ⇒ Object 



93
94
95
96
 
# File 'lib/kafka/cluster.rb', line 93

def refresh_metadata!
  @cluster_info = nil
  cluster_info
end

#refresh_metadata_if_necessary! ⇒ Object 



98
99
100
 
# File 'lib/kafka/cluster.rb', line 98

def refresh_metadata_if_necessary!
  refresh_metadata! if @stale
end

#resolve_offset(topic, partition, offset) ⇒ Object 



328
329
330
 
# File 'lib/kafka/cluster.rb', line 328

def resolve_offset(topic, partition, offset)
  resolve_offsets(topic, [partition], offset).fetch(partition)
end

#resolve_offsets(topic, partitions, offset) ⇒ Object 



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
 
# File 'lib/kafka/cluster.rb', line 285

def resolve_offsets(topic, partitions, offset)
  add_target_topics([topic])
  refresh_metadata_if_necessary!

  partitions_by_broker = partitions.each_with_object({}) {|partition, hsh|
    broker = get_leader(topic, partition)

    hsh[broker] ||= []
    hsh[broker] << partition
  }

  if offset == :earliest
    offset = -2
  elsif offset == :latest
    offset = -1
  end

  offsets = {}

  partitions_by_broker.each do |broker, broker_partitions|
    response = broker.list_offsets(
      topics: {
        topic => broker_partitions.map {|partition|
          {
            partition: partition,
            time: offset,
            max_offsets: 1,
          }
        }
      }
    )

    broker_partitions.each do |partition|
      offsets[partition] = response.offset_for(topic, partition)
    end
  end

  offsets
rescue Kafka::ProtocolError
  mark_as_stale!
  raise
end

#supports_api?(api_key, version = nil) ⇒ Boolean

Returns:

  • (Boolean) 


58
59
60
61
62
63
64
65
66
67
 
# File 'lib/kafka/cluster.rb', line 58

def supports_api?(api_key, version = nil)
  info = api_info(api_key)
  if info.nil?
    return false
  elsif version.nil?
    return true
  else
    return info.version_supported?(version)
  end
end

#topics ⇒ Object 



332
333
334
335
336
337
 
# File 'lib/kafka/cluster.rb', line 332

def topics
  refresh_metadata_if_necessary!
  cluster_info.topics.select do |topic|
    topic.topic_error_code == 0
  end.map(&:topic_name)
end