karafka 2.1.3 → 2.1.5.beta1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 6e65b59d084427bababbe797da6aa39ce76d04ceb235bdc43d5f85649e030c0f
-  data.tar.gz: a0599693bea1a05f2066180492cdb5017ac609b118195c64983990550f8e7e98
+  metadata.gz: 967412d28c31483df6a0c151ec0578367da24e503184608789791774347ffc53
+  data.tar.gz: '0828ba1db27f9f287b3ab9f80928816a81ae45ad79ce092e02cccd33b9bac204'
 SHA512:
-  metadata.gz: 9ef5dfbce059aedad3c24911a02e05068122d78d6e54f183c9b1155f1bb6b9f80cfc49cc7cdcc678ea427f86b1832f161aa7c9a411afbb2e6d83fc87bc67ffbe
-  data.tar.gz: 43f0a4ae3c9247df214d6fccd7ee9735a975b45cb0db49dd9490d675abe5d03ffeddbafc658d3953a39243304fd7754a265d60450d941f5a19dfa3cbe3716ac9
+  metadata.gz: 3774daedd74efe8e2498e97e503226cc01848e7a7be6ca255852bcaf112bc790ff98953584c771de8afa96142e8244e9f1b036b18490305766b428671b136bce
+  data.tar.gz: b575cf2be3d4cdcb73fec9cdfd65cd705408d1efae39ed200ae2d815e7c9241f582d5ebb8b802347289de910ee50589f0551bb4c194a161905e38425d05bd296

data/CHANGELOG.md

@@ -1,5 +1,17 @@
 # Karafka framework changelog
 
+## 2.1.5 (Unreleased)
+- [Improvement] Drastically improve `#revoked?` response quality by checking the real time assignment lost state on librdkafka.
+- [Improvement] Improve eviction of saturated jobs that would run on already revoked assignments.
+- [Improvement] Expose `#commit_offsets` and `#commit_offsets!` methods in the consumer to provide ability to commit offsets directly to Kafka without having to mark new messages as consumed.
+- [Improvement] No longer skip offset commit when no messages marked as consumed as `librdkafka` has fixed the crashes there.
+- [Improvement] Remove no longer needed patches.
+- [Change] Require `karafka-core` `>= 2.1.0`
+- [Change] Require `waterdrop` `>= 2.6.1`
+
+## 2.1.4 (2023-06-06)
+- [Fix] `processing_lag` and `consumption_lag` on empty batch fail on shutdown usage (#1475)
+
 ## 2.1.3 (2023-05-29)
 - [Maintenance] Add linter to ensure, that all integration specs end with `_spec.rb`.
 - [Fix] Fix `#retrying?` helper result value (Aerdayne).

data/Gemfile.lock

@@ -1,19 +1,19 @@
 PATH
   remote: .
   specs:
-    karafka (2.1.3)
-      karafka-core (>= 2.0.13, < 3.0.0)
+    karafka (2.1.5.beta1)
+      karafka-core (>= 2.1.0.beta1, < 2.2.0)
       thor (>= 0.20)
-      waterdrop (>= 2.5.3, < 3.0.0)
+      waterdrop (>= 2.6.1.beta1, < 3.0.0)
       zeitwerk (~> 2.3)
 
 GEM
   remote: https://rubygems.org/
   specs:
-    activejob (7.0.4.3)
-      activesupport (= 7.0.4.3)
+    activejob (7.0.5)
+      activesupport (= 7.0.5)
       globalid (>= 0.3.6)
-    activesupport (7.0.4.3)
+    activesupport (7.0.5)
       concurrent-ruby (~> 1.0, >= 1.0.2)
       i18n (>= 1.6, < 2)
       minitest (>= 5.1)
@@ -28,26 +28,26 @@ GEM
     ffi (1.15.5)
     globalid (1.1.0)
       activesupport (>= 5.0)
-    i18n (1.13.0)
+    i18n (1.14.1)
       concurrent-ruby (~> 1.0)
-    karafka-core (2.0.13)
+    karafka-core (2.1.0.beta1)
       concurrent-ruby (>= 1.1)
-      karafka-rdkafka (>= 0.12.3)
-    karafka-rdkafka (0.12.3)
+      karafka-rdkafka (>= 0.13.0.beta2, < 0.14.0)
+    karafka-rdkafka (0.13.0.beta2)
       ffi (~> 1.15)
       mini_portile2 (~> 2.6)
       rake (> 12)
-    karafka-web (0.5.2)
+    karafka-web (0.6.0)
       erubi (~> 1.4)
-      karafka (>= 2.0.40, < 3.0.0)
-      karafka-core (>= 2.0.12, < 3.0.0)
-      roda (~> 3.63)
+      karafka (>= 2.1.4, < 3.0.0)
+      karafka-core (>= 2.0.13, < 3.0.0)
+      roda (~> 3.68, >= 3.68)
       tilt (~> 2.0)
     mini_portile2 (2.8.2)
-    minitest (5.18.0)
-    rack (3.0.7)
+    minitest (5.18.1)
+    rack (3.0.8)
     rake (13.0.6)
-    roda (3.68.0)
+    roda (3.69.0)
       rack
     rspec (3.12.0)
       rspec-core (~> 3.12.0)
@@ -69,11 +69,11 @@ GEM
     simplecov-html (0.12.3)
     simplecov_json_formatter (0.1.4)
     thor (1.2.2)
-    tilt (2.1.0)
+    tilt (2.2.0)
     tzinfo (2.0.6)
       concurrent-ruby (~> 1.0)
-    waterdrop (2.5.3)
-      karafka-core (>= 2.0.13, < 3.0.0)
+    waterdrop (2.6.1.beta1)
+      karafka-core (>= 2.1.0.beta1, < 3.0.0)
       zeitwerk (~> 2.3)
     zeitwerk (2.6.8)
 

data/karafka.gemspec

@@ -21,9 +21,9 @@ Gem::Specification.new do |spec|
     without having to focus on things that are not your business domain.
   DESC
 
-  spec.add_dependency 'karafka-core', '>= 2.0.13', '< 3.0.0'
+  spec.add_dependency 'karafka-core', '>= 2.1.0.beta1', '< 2.2.0'
   spec.add_dependency 'thor', '>= 0.20'
-  spec.add_dependency 'waterdrop', '>= 2.5.3', '< 3.0.0'
+  spec.add_dependency 'waterdrop', '>= 2.6.1.beta1', '< 3.0.0'
   spec.add_dependency 'zeitwerk', '~> 2.3'
 
   if $PROGRAM_NAME.end_with?('gem')

data/lib/karafka/admin.rb

@@ -171,7 +171,9 @@ module Karafka
       # @return [Rdkafka::Metadata] cluster metadata info
       def cluster_info
         with_admin do |admin|
-          Rdkafka::Metadata.new(admin.instance_variable_get('@native_kafka'))
+          admin.instance_variable_get('@native_kafka').with_inner do |inner|
+            Rdkafka::Metadata.new(inner)
+          end
         end
       end
 

data/lib/karafka/base_consumer.rb

@@ -215,10 +215,13 @@ module Karafka
     end
 
     # @return [Boolean] true if partition was revoked from the current consumer
-    # @note We know that partition got revoked because when we try to mark message as consumed,
-    #   unless if is successful, it will return false
+    # @note There are two "levels" on which we can know that partition was revoked. First one is
+    #   when we loose the assignment involuntarily and second is when coordinator gets this info
+    #   after we poll with the rebalance callbacks. The first check allows us to get this notion
+    #   even before we poll but it gets reset when polling happens, hence we also need to switch
+    #   the coordinator state after the revocation (but prior to running more jobs)
     def revoked?
-      coordinator.revoked?
+      client.assignment_lost? || coordinator.revoked?
     end
 
     # @return [Boolean] are we retrying processing after an error. This can be used to provide a

data/lib/karafka/connection/client.rb

@@ -41,9 +41,6 @@ module Karafka
         @buffer = RawMessagesBuffer.new
         @rebalance_manager = RebalanceManager.new
         @kafka = build_consumer
-        # Marks if we need to offset. If we did not store offsets, we should not commit the offset
-        # position as it will crash rdkafka
-        @offsetting = false
         # We need to keep track of what we have paused for resuming
         # In case we loose partition, we still need to resume it, otherwise it won't be fetched
         # again if we get reassigned to it later on. We need to keep them as after revocation we
@@ -109,8 +106,14 @@ module Karafka
         end
       end
 
+      # @return [Boolean] true if our current assignment has been lost involuntarily.
+      def assignment_lost?
+        @mutex.synchronize do
+          @kafka.assignment_lost?
+        end
+      end
+
       # Commits the offset on a current consumer in a non-blocking or blocking way.
-      # Ignoring a case where there would not be an offset (for example when rebalance occurs).
       #
       # @param async [Boolean] should the commit happen async or sync (async by default)
       # @return [Boolean] did committing was successful. It may be not, when we no longer own
@@ -118,7 +121,7 @@ module Karafka
       #
       # @note This will commit all the offsets for the whole consumer. In order to achieve
       #   granular control over where the offset should be for particular topic partitions, the
-      #   store_offset should be used to only store new offset when we want to to be flushed
+      #   store_offset should be used to only store new offset when we want them to be flushed
       def commit_offsets(async: true)
         @mutex.lock
 
@@ -256,7 +259,6 @@ module Karafka
 
         @mutex.synchronize do
           @closed = false
-          @offsetting = false
           @paused_tpls.clear
           @kafka = build_consumer
         end
@@ -281,7 +283,6 @@ module Karafka
       # @param message [Karafka::Messages::Message]
       # @return [Boolean] true if we could store the offset (if we still own the partition)
       def internal_store_offset(message)
-        @offsetting = true
         @kafka.store_offset(message)
         true
       rescue Rdkafka::RdkafkaError => e
@@ -294,11 +295,11 @@ module Karafka
       # Non thread-safe message committing method
       # @param async [Boolean] should the commit happen async or sync (async by default)
       # @return [Boolean] true if offset commit worked, false if we've lost the assignment
+      # @note We do **not** consider `no_offset` as any problem and we allow to commit offsets
+      #   even when no stored, because with sync commit, it refreshes the ownership state of the
+      #   consumer in a sync way.
       def internal_commit_offsets(async: true)
-        return true unless @offsetting
-
         @kafka.commit(nil, async)
-        @offsetting = false
 
         true
       rescue Rdkafka::RdkafkaError => e

data/lib/karafka/connection/rebalance_manager.rb

@@ -49,9 +49,8 @@ module Karafka
       # Callback that kicks in inside of rdkafka, when new partitions are assigned.
       #
       # @private
-      # @param _ [Rdkafka::Consumer]
       # @param partitions [Rdkafka::Consumer::TopicPartitionList]
-      def on_partitions_assigned(_, partitions)
+      def on_partitions_assigned(partitions)
         @assigned_partitions = partitions.to_h.transform_values { |part| part.map(&:partition) }
         @changed = true
       end
@@ -59,9 +58,8 @@ module Karafka
       # Callback that kicks in inside of rdkafka, when partitions are revoked.
       #
       # @private
-      # @param _ [Rdkafka::Consumer]
       # @param partitions [Rdkafka::Consumer::TopicPartitionList]
-      def on_partitions_revoked(_, partitions)
+      def on_partitions_revoked(partitions)
         @revoked_partitions = partitions.to_h.transform_values { |part| part.map(&:partition) }
         @changed = true
       end

data/lib/karafka/messages/batch_metadata.rb

@@ -20,14 +20,21 @@ module Karafka
     ) do
       # This lag describes how long did it take for a message to be consumed from the moment it was
       # created
+      #
+      #
+      # @return [Integer] number of milliseconds
+      # @note In case of usage in workless flows, this value will be set to -1
       def consumption_lag
-        time_distance_in_ms(processed_at, created_at)
+        processed_at ? time_distance_in_ms(processed_at, created_at) : -1
       end
 
       # This lag describes how long did a batch have to wait before it was picked up by one of the
       # workers
+      #
+      # @return [Integer] number of milliseconds
+      # @note In case of usage in workless flows, this value will be set to -1
       def processing_lag
-        time_distance_in_ms(processed_at, scheduled_at)
+        processed_at ? time_distance_in_ms(processed_at, scheduled_at) : -1
       end
 
       private

data/lib/karafka/messages/builders/message.rb

@@ -12,9 +12,6 @@ module Karafka
           # @param received_at [Time] moment when we've received the message
           # @return [Karafka::Messages::Message] message object with payload and metadata
           def call(kafka_message, topic, received_at)
-            # @see https://github.com/appsignal/rdkafka-ruby/issues/168
-            kafka_message.headers.transform_keys!(&:to_s)
-
             metadata = Karafka::Messages::Metadata.new(
               timestamp: kafka_message.timestamp,
               headers: kafka_message.headers,

data/lib/karafka/patches/rdkafka/bindings.rb

@@ -49,14 +49,13 @@ module Karafka
           #
           # @param code [Integer]
           # @param opaque [Rdkafka::Opaque]
-          # @param consumer [Rdkafka::Consumer]
           # @param tpl [Rdkafka::Consumer::TopicPartitionList]
-          def trigger_callbacks(code, opaque, consumer, tpl)
+          def trigger_callbacks(code, opaque, tpl)
             case code
             when RB::RD_KAFKA_RESP_ERR__ASSIGN_PARTITIONS
-              opaque.call_on_partitions_assigned(consumer, tpl)
+              opaque.call_on_partitions_assigned(tpl)
             when RB::RD_KAFKA_RESP_ERR__REVOKE_PARTITIONS
-              opaque.call_on_partitions_revoked(consumer, tpl)
+              opaque.call_on_partitions_revoked(tpl)
             end
           rescue StandardError => e
             Karafka.monitor.instrument(
@@ -92,9 +91,8 @@ module Karafka
           return unless opaque
 
           tpl = ::Rdkafka::Consumer::TopicPartitionList.from_native_tpl(partitions_ptr).freeze
-          consumer = ::Rdkafka::Consumer.new(client_ptr)
 
-          pr.trigger_callbacks(code, opaque, consumer, tpl)
+          pr.trigger_callbacks(code, opaque, tpl)
         end
       end
     end

data/lib/karafka/processing/strategies/default.rb

@@ -60,6 +60,24 @@ module Karafka
           true
         end
 
+        # Triggers an async offset commit
+        #
+        # @return [Boolean] true if we still own the partition.
+        # @note Due to its async nature, this may not fully represent the offset state in some
+        #   edge cases (like for example going beyond max.poll.interval)
+        def commit_offsets
+          client.commit_offsets(async: true)
+        end
+
+        # Triggers a synchronous offsets commit to Kafka
+        #
+        # @return [Boolean] true if we still own the partition, false otherwise.
+        # @note This is fully synchronous, hence the result of this can be used in DB transactions
+        #   etc as a way of making sure, that we still own the partition.
+        def commit_offsets!
+          client.commit_offsets(async: false)
+        end
+
         # No actions needed for the standard flow here
         def handle_before_enqueue
           nil

data/lib/karafka/setup/attributes_map.rb

@@ -47,6 +47,7 @@ module Karafka
         fetch.max.bytes
         fetch.message.max.bytes
         fetch.min.bytes
+        fetch.queue.backoff.ms
         fetch.wait.max.ms
         group.id
         group.instance.id

data/lib/karafka/version.rb

@@ -3,5 +3,5 @@
 # Main module namespace
 module Karafka
   # Current Karafka version
-  VERSION = '2.1.3'
+  VERSION = '2.1.5.beta1'
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: karafka
 version: !ruby/object:Gem::Version
-  version: 2.1.3
+  version: 2.1.5.beta1
 platform: ruby
 authors:
 - Maciej Mensfeld
@@ -35,7 +35,7 @@ cert_chain:
   Qf04B9ceLUaC4fPVEz10FyobjaFoY4i32xRto3XnrzeAgfEe4swLq8bQsR3w/EF3
   MGU0FeSV2Yj7Xc2x/7BzLK8xQn5l7Yy75iPF+KP3vVmDHnNl
   -----END CERTIFICATE-----
-date: 2023-05-29 00:00:00.000000000 Z
+date: 2023-06-18 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: karafka-core
@@ -43,20 +43,20 @@ dependencies:
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: 2.0.13
+        version: 2.1.0.beta1
     - - "<"
       - !ruby/object:Gem::Version
-        version: 3.0.0
+        version: 2.2.0
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: 2.0.13
+        version: 2.1.0.beta1
     - - "<"
       - !ruby/object:Gem::Version
-        version: 3.0.0
+        version: 2.2.0
 - !ruby/object:Gem::Dependency
   name: thor
   requirement: !ruby/object:Gem::Requirement
@@ -77,7 +77,7 @@ dependencies:
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: 2.5.3
+        version: 2.6.1.beta1
     - - "<"
       - !ruby/object:Gem::Version
         version: 3.0.0
@@ -87,7 +87,7 @@ dependencies:
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: 2.5.3
+        version: 2.6.1.beta1
     - - "<"
       - !ruby/object:Gem::Version
         version: 3.0.0
@@ -214,7 +214,6 @@ files:
 - lib/karafka/messages/parser.rb
 - lib/karafka/messages/seek.rb
 - lib/karafka/patches/rdkafka/bindings.rb
-- lib/karafka/patches/rdkafka/consumer.rb
 - lib/karafka/pro.rb
 - lib/karafka/pro/active_job/consumer.rb
 - lib/karafka/pro/active_job/dispatcher.rb
@@ -415,9 +414,9 @@ required_ruby_version: !ruby/object:Gem::Requirement
       version: '0'
 required_rubygems_version: !ruby/object:Gem::Requirement
   requirements:
-  - - ">="
+  - - ">"
     - !ruby/object:Gem::Version
-      version: '0'
+      version: 1.3.1
 requirements: []
 rubygems_version: 3.4.10
 signing_key: 

data/lib/karafka/patches/rdkafka/consumer.rb

@@ -1,22 +0,0 @@
-# frozen_string_literal: true
-
-module Karafka
-  # Patches to external components
-  module Patches
-    # Rdkafka related patches
-    module Rdkafka
-      # Rdkafka::Consumer patches
-      module Consumer
-        # A method that allows us to get the native kafka producer name
-        # @return [String] producer instance name
-        # @note We need this to make sure that we allocate proper dispatched events only to
-        #   callback listeners that should publish them
-        def name
-          @name ||= ::Rdkafka::Bindings.rd_kafka_name(@native_kafka)
-        end
-      end
-    end
-  end
-end
-
-::Rdkafka::Consumer.include ::Karafka::Patches::Rdkafka::Consumer