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

#clear_target_topics ⇒ nil

Clears the list of target topics.

Returns:

  • (nil)

See Also:



56
57
58
59
 
# File 'lib/kafka/cluster.rb', line 56

def clear_target_topics
  @target_topics.clear
  refresh_metadata!
end

#disconnect ⇒ Object 



184
185
186
 
# File 'lib/kafka/cluster.rb', line 184

def disconnect
  @broker_pool.close
end

#get_group_coordinator(group_id:) ⇒ Object

Raises:



83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
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
 
# File 'lib/kafka/cluster.rb', line 83

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.



79
80
81
 
# File 'lib/kafka/cluster.rb', line 79

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

#mark_as_stale! ⇒ Object 



61
62
63
 
# File 'lib/kafka/cluster.rb', line 61

def mark_as_stale!
  @stale = true
end

#partitions_for(topic) ⇒ Object 



124
125
126
127
128
129
130
131
 
# File 'lib/kafka/cluster.rb', line 124

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 



65
66
67
68
 
# File 'lib/kafka/cluster.rb', line 65

def refresh_metadata!
  @cluster_info = nil
  cluster_info
end

#refresh_metadata_if_necessary! ⇒ Object 



70
71
72
 
# File 'lib/kafka/cluster.rb', line 70

def refresh_metadata_if_necessary!
  refresh_metadata! if @stale
end

#resolve_offset(topic, partition, offset) ⇒ Object 



176
177
178
 
# File 'lib/kafka/cluster.rb', line 176

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

#resolve_offsets(topic, partitions, offset) ⇒ Object 



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
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
 
# File 'lib/kafka/cluster.rb', line 133

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 



180
181
182
 
# File 'lib/kafka/cluster.rb', line 180

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