Class: Karafka::Processing::Executor

Inherits:
Object
  • Object
show all
Extended by:
Forwardable
Defined in:
lib/karafka/processing/executor.rb

Overview

Note:

Executors are not removed after partition is revoked. They are not that big and will be re-used in case of a re-claim

Executors:

  • run consumers code (for ‘#call`) or run given preparation / teardown operations when needed from separate threads.

  • they re-create consumer instances in case of partitions that were revoked and assigned back.

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(group_id, client, coordinator) ⇒ Executor

Returns a new instance of Executor.

Parameters:



34
35
36
37
38
39
 
# File 'lib/karafka/processing/executor.rb', line 34

def initialize(group_id, client, coordinator)
  @id = SecureRandom.hex(6)
  @group_id = group_id
  @client = client
  @coordinator = coordinator
end

Instance Attribute Details

#coordinator ⇒ Karafka::Processing::Coordinator (readonly)

Returns coordinator for this executor.

Returns:



29
30
31
 
# File 'lib/karafka/processing/executor.rb', line 29

def coordinator
  @coordinator
end

#group_id ⇒ String (readonly)

Returns subscription group id to which a given executor belongs.

Returns:

  • (String) —

    subscription group id to which a given executor belongs



23
24
25
 
# File 'lib/karafka/processing/executor.rb', line 23

def group_id
  @group_id
end

#id ⇒ String (readonly)

Returns unique id that we use to ensure, that we use for state tracking.

Returns:

  • (String) —

    unique id that we use to ensure, that we use for state tracking



20
21
22
 
# File 'lib/karafka/processing/executor.rb', line 20

def id
  @id
end

#messages ⇒ Karafka::Messages::Messages (readonly)

Returns messages batch.

Returns:



26
27
28
 
# File 'lib/karafka/processing/executor.rb', line 26

def messages
  @messages
end

Instance Method Details

#after_consume ⇒ Object

Runs consumer after consumption code



78
79
80
 
# File 'lib/karafka/processing/executor.rb', line 78

def after_consume
  consumer.on_after_consume
end

#before_consume ⇒ Object

Runs setup and warm-up code in the worker prior to running the consumption



67
68
69
 
# File 'lib/karafka/processing/executor.rb', line 67

def before_consume
  consumer.on_before_consume
end

#before_enqueue(messages) ⇒ Object

Allows us to prepare the consumer in the listener thread prior to the job being send to the queue. It also allows to run some code that is time sensitive and cannot wait in the queue as it could cause starvation.

Parameters:



46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
 
# File 'lib/karafka/processing/executor.rb', line 46

def before_enqueue(messages)
  # Recreate consumer with each batch if persistence is not enabled
  # We reload the consumers with each batch instead of relying on some external signals
  # when needed for consistency. That way devs may have it on or off and not in this
  # middle state, where re-creation of a consumer instance would occur only sometimes
  @consumer = nil unless topic.consumer_persistence

  # First we build messages batch...
  consumer.messages = Messages::Builders::Messages.call(
    messages,
    topic,
    partition,
    # the moment we've received the batch or actually the moment we've enqueued it,
    # but good enough
    Time.now
  )

  consumer.on_before_enqueue
end

#consume ⇒ Object

Runs consumer data processing against given batch and handles failures and errors.



72
73
74
75
 
# File 'lib/karafka/processing/executor.rb', line 72

def consume
  # We run the consumer client logic...
  consumer.on_consume
end

#idle ⇒ Object

Runs consumer idle operations This may include house-keeping or other state management changes that can occur but that not mean there are any new messages available for the end user to process



85
86
87
88
89
90
91
92
93
94
95
96
97
 
# File 'lib/karafka/processing/executor.rb', line 85

def idle
  # Initializes the messages set in case idle operation would happen before any processing
  # This prevents us from having no messages object at all as the messages object and
  # its metadata may be used for statistics
  consumer.messages ||= Messages::Builders::Messages.call(
    [],
    topic,
    partition,
    Time.now
  )

  consumer.on_idle
end

#revoked ⇒ Object

Note:

Clearing the consumer will ensure, that if we get the partition back, it will be handled with a consumer with a clean state.

Note:

We run it only when consumer was present, because presence indicates, that at least a single message has been consumed.

Note:

We do not reset the consumer but we indicate need for recreation instead, because after the revocation, there still may be ‘#after_consume` running that needs a given consumer instance.

Runs the controller ‘#revoked` method that should be triggered when a given consumer is no longer needed due to partitions reassignment.



111
112
113
 
# File 'lib/karafka/processing/executor.rb', line 111

def revoked
  consumer.on_revoked if @consumer
end

#shutdown ⇒ Object

Note:

While we do not need to clear the consumer here, it’s a good habit to clean after work is done.

Runs the controller ‘#shutdown` method that should be triggered when a given consumer is no longer needed as we’re closing the process.



120
121
122
123
124
 
# File 'lib/karafka/processing/executor.rb', line 120

def shutdown
  # There is a case, where the consumer no longer exists because it was revoked, in case like
  # that we do not build a new instance and shutdown should not be triggered.
  consumer.on_shutdown if @consumer
end