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:



19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
 
# File 'lib/kafka/cluster.rb', line 19

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) 


40
41
42
43
44
45
46
47
48
49
50
 
# File 'lib/kafka/cluster.rb', line 40

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

#api_info(api_key) ⇒ Object 



52
53
54
 
# File 'lib/kafka/cluster.rb', line 52

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

#apis ⇒ Object 



56
57
58
59
60
61
62
63
64
65
 
# File 'lib/kafka/cluster.rb', line 56

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:



71
72
73
74
 
# File 'lib/kafka/cluster.rb', line 71

def clear_target_topics
  @target_topics.clear
  refresh_metadata!
end

#create_topic(name, num_partitions: 1, replication_factor: 1, timeout: 30) ⇒ Object 



148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
 
# File 'lib/kafka/cluster.rb', line 148

def create_topic(name, num_partitions: 1, replication_factor: 1, timeout: 30)
  options = {
    topics: {
      name => {
        num_partitions: num_partitions,
        replication_factor: replication_factor,
      }
    },
    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
  end

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

#disconnect ⇒ Object 



235
236
237
 
# File 'lib/kafka/cluster.rb', line 235

def disconnect
  @broker_pool.close
end

#get_group_coordinator(group_id:) ⇒ Object

Raises:



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

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.



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

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

#mark_as_stale! ⇒ Object 



76
77
78
 
# File 'lib/kafka/cluster.rb', line 76

def mark_as_stale!
  @stale = true
end

#partitions_for(topic) ⇒ Object 



139
140
141
142
143
144
145
146
 
# File 'lib/kafka/cluster.rb', line 139

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 



80
81
82
83
 
# File 'lib/kafka/cluster.rb', line 80

def refresh_metadata!
  @cluster_info = nil
  cluster_info
end

#refresh_metadata_if_necessary! ⇒ Object 



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

def refresh_metadata_if_necessary!
  refresh_metadata! if @stale
end

#resolve_offset(topic, partition, offset) ⇒ Object 



226
227
228
 
# File 'lib/kafka/cluster.rb', line 226

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

#resolve_offsets(topic, partitions, offset) ⇒ Object 



183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
 
# File 'lib/kafka/cluster.rb', line 183

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

#topics ⇒ Object 



230
231
232
233
 
# File 'lib/kafka/cluster.rb', line 230

def topics
  refresh_metadata_if_necessary!
  cluster_info.topics.map(&:topic_name)
end