waterdrop 2.8.12 → 2.8.13

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: a0890fcd73147d293340b55cca58c975d81726208c609be1f338ce56d87b4d18
-  data.tar.gz: 7a2743c068af9936186fa852d6e663c85c172aa93de76b16aaf09e2c1de24a25
+  metadata.gz: da869aec5fa217a2adda674c787dd142a4f4e50e99a9b86d349d42f700821c88
+  data.tar.gz: 1934d9b60db038b6ea2f99faf7961a5f3ea2b3dfe845f84632db7f145f094cd2
 SHA512:
-  metadata.gz: 66136311cb010a1648a0a5440cd930fcb96bb75fe30575d5c14effb8550a7cea4295dda98828790ac567ec9f3346dde7901c11228f6ca251fdf454a5354fad9d
-  data.tar.gz: a701a63fe3a171084b63ac46ca9f66d75572b7efbfed0b794884df0fdb2b272404a6594ba239f16956cdc9c5fbecb31dbb5574fc0e5c1f12225ce61514c776a3
+  metadata.gz: c25b0221d070e1ba2b9badcff468caad02d4f4da158083a3514aa8a4d2b5b867579d6bebdaa35f45416c97de6bdddcb9c080a9d5399bca7f509e26e9f1775140
+  data.tar.gz: d2c32960c72034b00a48d3cd5b7ab30e12908638945f08aa69ddd27947b98cf38b9b3b0d3ccf78b02ce25b569a5ad551d1bedcbaf40b06df0b1caa7565567b56

data/.github/workflows/ci.yml

@@ -17,7 +17,6 @@ jobs:
   specs:
     timeout-minutes: 15
     runs-on: ubuntu-latest
-    needs: diffend
     env:
       BUNDLE_FORCE_RUBY_PLATFORM: ${{ matrix.force_ruby_platform }}
     strategy:
@@ -87,29 +86,6 @@ jobs:
       - name: Check test topics naming convention
         run: bin/verify_topics_naming
 
-  diffend:
-    timeout-minutes: 5
-
-    runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
-    steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-        with:
-          fetch-depth: 0
-      - name: Set up Ruby
-        uses: ruby/setup-ruby@2a7b30092b0caf9c046252510f9273b4875f3db9 # v1.254.0
-        with:
-          ruby-version: 3.4
-          self-hosted: false
-
-      - name: Install latest bundler
-        run: gem install bundler --no-document
-      - name: Install Diffend plugin
-        run: bundle plugin install diffend
-      - name: Bundle Secure
-        run: bundle secure
-
   coditsu:
     timeout-minutes: 5
     runs-on: ubuntu-latest
@@ -139,7 +115,6 @@ jobs:
     runs-on: ubuntu-latest
     if: always()
     needs:
-      - diffend
       - coditsu
       - specs
     steps:

data/CHANGELOG.md

@@ -1,5 +1,10 @@
 # WaterDrop changelog
 
+## 2.8.13 (2025-10-31)
+- [Enhancement] Make `fenced` error skip-reload behavior configurable via new `non_reloadable_errors` setting (defaults to `[:fenced]` for backward compatibility).
+- [Enhancement] Add `producer.reload` event allowing config modification before reload to escape fencing loops (#706).
+- [Enhancement] Do not early initialize the new instance on reload.
+
 ## 2.8.12 (2025-10-10)
 - [Enhancement] Introduce `reload_on_idempotent_fatal_error` to automatically reload librdkafka producer after fatal errors on idempotent (non-transactional) producers.
 - [Enhancement] Add configurable backoff and retry limits for fatal error recovery to prevent infinite reload loops:
@@ -19,7 +24,7 @@
 ## 2.8.11 (2025-09-27)
 - [Enhancement] Provide fast-track for middleware-less flows (20% faster) for single message, 5000x faster for batches.
 - [Enhancement] Optimize middlewares application by around 20%.
-- [Change] Remove Ruby `3.1` according to the EOL schedule.
+- **[EOL]** Remove Ruby `3.1` according to the EOL schedule.
 - [Fix] Connection pool timeout parameter now accepts milliseconds instead of seconds for consistency with other WaterDrop timeouts. The default timeout has been changed from `5` seconds to `5000` milliseconds (equivalent value).
 
 ## 2.8.10 (2025-09-25)
@@ -33,7 +38,7 @@
 ## 2.8.8 (2025-09-23)
 - [Feature] Add `WaterDrop::ConnectionPool` for efficient connection pooling using the proven `connection_pool` gem.
 - [Feature] Add `WaterDrop.instrumentation` class-level instrumentation for producer lifecycle events. This allows external libraries to subscribe to `producer.created` and `producer.configured` events without needing producer instance references, enabling middleware injection and configuration by libraries like Datadog tracing.
-- [Change] Remove Ruby `3.1` specs according to the EOL schedule.
+- **[EOL]** Remove Ruby `3.1` specs according to the EOL schedule.
 
 ## 2.8.7 (2025-09-02)
 - [Enhancement] Disable Nagle algorithm by default (improves latency / aligned with librdkafka)

data/Gemfile

@@ -2,12 +2,10 @@
 
 source 'https://rubygems.org'
 
-plugin 'diffend'
-
 gemspec
 
 # Relaxed from 2.7 because we support Ruby 3.1
-gem 'zeitwerk', '~> 2.6.18'
+gem 'zeitwerk', '~> 2.7.0'
 
 group :development do
   gem 'byebug'

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    waterdrop (2.8.12)
+    waterdrop (2.8.13)
       karafka-core (>= 2.4.9, < 3.0.0)
       karafka-rdkafka (>= 0.20.0)
       zeitwerk (~> 2.3)
@@ -24,35 +24,35 @@ GEM
     ffi (1.17.2-x86_64-darwin)
     ffi (1.17.2-x86_64-linux-gnu)
     ffi (1.17.2-x86_64-linux-musl)
-    json (2.13.2)
-    karafka-core (2.5.6)
+    json (2.15.1)
+    karafka-core (2.5.7)
       karafka-rdkafka (>= 0.20.0)
       logger (>= 1.6.0)
-    karafka-rdkafka (0.21.0)
+    karafka-rdkafka (0.22.2)
       ffi (~> 1.15)
       json (> 2.0)
       logger
       mini_portile2 (~> 2.6)
       rake (> 12)
-    karafka-rdkafka (0.21.0-aarch64-linux-gnu)
+    karafka-rdkafka (0.22.2-aarch64-linux-gnu)
       ffi (~> 1.15)
       json (> 2.0)
       logger
       mini_portile2 (~> 2.6)
       rake (> 12)
-    karafka-rdkafka (0.21.0-arm64-darwin)
+    karafka-rdkafka (0.22.2-arm64-darwin)
       ffi (~> 1.15)
       json (> 2.0)
       logger
       mini_portile2 (~> 2.6)
       rake (> 12)
-    karafka-rdkafka (0.21.0-x86_64-linux-gnu)
+    karafka-rdkafka (0.22.2-x86_64-linux-gnu)
       ffi (~> 1.15)
       json (> 2.0)
       logger
       mini_portile2 (~> 2.6)
       rake (> 12)
-    karafka-rdkafka (0.21.0-x86_64-linux-musl)
+    karafka-rdkafka (0.22.2-x86_64-linux-musl)
       ffi (~> 1.15)
       json (> 2.0)
       logger
@@ -62,27 +62,27 @@ GEM
     mini_portile2 (2.8.9)
     ostruct (0.6.3)
     rake (13.3.0)
-    rspec (3.13.1)
+    rspec (3.13.2)
       rspec-core (~> 3.13.0)
       rspec-expectations (~> 3.13.0)
       rspec-mocks (~> 3.13.0)
-    rspec-core (3.13.4)
+    rspec-core (3.13.6)
       rspec-support (~> 3.13.0)
     rspec-expectations (3.13.5)
       diff-lcs (>= 1.2.0, < 2.0)
       rspec-support (~> 3.13.0)
-    rspec-mocks (3.13.5)
+    rspec-mocks (3.13.6)
       diff-lcs (>= 1.2.0, < 2.0)
       rspec-support (~> 3.13.0)
-    rspec-support (3.13.4)
+    rspec-support (3.13.6)
     simplecov (0.22.0)
       docile (~> 1.1)
       simplecov-html (~> 0.11)
       simplecov_json_formatter (~> 0.1)
-    simplecov-html (0.13.1)
+    simplecov-html (0.13.2)
     simplecov_json_formatter (0.1.4)
     warning (1.5.0)
-    zeitwerk (2.6.18)
+    zeitwerk (2.7.3)
 
 PLATFORMS
   aarch64-linux-gnu
@@ -105,7 +105,7 @@ DEPENDENCIES
   simplecov
   warning
   waterdrop!
-  zeitwerk (~> 2.6.18)
+  zeitwerk (~> 2.7.0)
 
 BUNDLED WITH
    2.7.0

data/config/locales/errors.yml

@@ -23,6 +23,7 @@ en:
       reload_on_transaction_fatal_error_format: must be boolean
       wait_backoff_on_transaction_fatal_error_format: must be a numeric that is equal or bigger to 0
       max_attempts_on_transaction_fatal_error_format: must be an integer that is equal or bigger than 1
+      non_reloadable_errors_format: must be an array of symbols
       oauth.token_provider_listener_format: 'must be false or respond to #on_oauthbearer_token_refresh'
       idle_disconnect_timeout_format: 'must be an integer that is equal to 0 or bigger than 30 000 (30 seconds)'
 

data/lib/waterdrop/config.rb

@@ -95,6 +95,20 @@ module WaterDrop
     # option [Integer] How many times to attempt reloading on transactional fatal error before
     #   giving up. This prevents infinite reload loops if the producer never recovers.
     setting :max_attempts_on_transaction_fatal_error, default: 10
+    # option [Array<Symbol>] List of fatal error codes that should NOT trigger producer reload.
+    #   These errors represent states that cannot be recovered by simply recreating the client.
+    #
+    #   WARNING: Modifying this setting can cause infinite reload loops if not properly understood.
+    #   The default includes :fenced errors because:
+    #   - Fencing occurs when another producer with the same transactional.id takes over
+    #   - This is an unrecoverable state - reloading won't help as the other producer is active
+    #   - Attempting to reload on fenced errors creates: produce -> fenced -> reload -> produce ->
+    #     fenced -> reload (infinite loop)
+    #
+    #   Only remove :fenced from this list if you have explicit logic to handle producer fencing
+    #   in your application (e.g., coordinated transactional.id assignment, manual intervention).
+    #   In most cases, you should keep the default value.
+    setting :non_reloadable_errors, default: %i[fenced]
     # option [Integer] Idle disconnect timeout in milliseconds. When set to 0, idle disconnection
     #   is disabled. When set to a positive value, WaterDrop will automatically disconnect
     #   producers that haven't sent any messages for the specified time period. This helps preserve

data/lib/waterdrop/contracts/config.rb

@@ -30,6 +30,9 @@ module WaterDrop
       required(:max_attempts_on_idempotent_fatal_error) { |val| val.is_a?(Integer) && val >= 1 }
       required(:wait_backoff_on_transaction_fatal_error) { |val| val.is_a?(Numeric) && val >= 0 }
       required(:max_attempts_on_transaction_fatal_error) { |val| val.is_a?(Integer) && val >= 1 }
+      required(:non_reloadable_errors) do |val|
+        val.is_a?(Array) && val.all?(Symbol)
+      end
       required(:idle_disconnect_timeout) do |val|
         val.is_a?(Integer) && (val.zero? || val >= 30_000)
       end

data/lib/waterdrop/instrumentation/notifications.rb

@@ -10,6 +10,7 @@ module WaterDrop
         producer.connected
         producer.closing
         producer.closed
+        producer.reload
         producer.reloaded
         producer.disconnecting
         producer.disconnected

data/lib/waterdrop/producer/idempotence.rb

@@ -23,13 +23,13 @@ module WaterDrop
       #   - Producer is idempotent
       #   - Producer is not transactional
       #   - reload_on_idempotent_fatal_error config is enabled
-      #   - Error is not in the NON_RELOADABLE_FATAL_ERRORS list
+      #   - Error is not in the non_reloadable_errors config list
       def idempotent_reloadable?(error)
         return false unless error.fatal?
         return false unless idempotent?
         return false if transactional?
         return false unless config.reload_on_idempotent_fatal_error
-        return false if NON_RELOADABLE_FATAL_ERRORS.include?(error.code)
+        return false if config.non_reloadable_errors.include?(error.code)
 
         true
       end
@@ -50,21 +50,34 @@ module WaterDrop
       # old client, and create a new client instance to continue operations.
       #
       # @param attempt [Integer] the current reload attempt number
+      # @param error [Rdkafka::RdkafkaError] the error that triggered the reload
       #
       # @note This is only called for idempotent, non-transactional producers when
       #   `reload_on_idempotent_fatal_error` is enabled
       # @note After reload, the producer will automatically retry the failed operation
-      def idempotent_reload_client_on_fatal_error(attempt)
+      def idempotent_reload_client_on_fatal_error(attempt, error)
         @operating_mutex.synchronize do
+          # Emit producer.reload event before reload
+          # Users can subscribe to this event and modify event[:caller].config.kafka to change
+          # producer config
+          @monitor.instrument(
+            'producer.reload',
+            producer_id: id,
+            error: error,
+            attempt: attempt,
+            caller: self
+          )
+
+          # Clear cached state that depends on config
+          # We always clear @idempotent as it might have been modified via the event
+          @idempotent = nil
+
           @monitor.instrument(
             'producer.reloaded',
             producer_id: id,
             attempt: attempt
           ) do
-            @client.flush(current_variant.max_wait_timeout)
-            purge
-            @client.close
-            @client = Builder.new.call(self, @config)
+            reload!
           end
         end
       end

data/lib/waterdrop/producer/transactions.rb

@@ -281,24 +281,41 @@ module WaterDrop
 
         return unless rd_error.is_a?(Rdkafka::RdkafkaError)
         return unless config.reload_on_transaction_fatal_error
-        return if NON_RELOADABLE_FATAL_ERRORS.include?(rd_error.code)
+        return if config.non_reloadable_errors.include?(rd_error.code)
 
         # Check if we've exceeded max reload attempts
         return unless transactional_retryable?
+        # We bubble up transactional errors, so there are cases where when fencing is not
+        # considered a non-reloadable, two layers of error handling would attempt to reload the
+        # client causing double reload. This halts reload if we're in a configured state as it
+        # means, we've already reloaded and we are not even yet connected
+        return if @status.configured?
 
         # Increment attempts before reload
         @transaction_fatal_error_attempts += 1
 
         @operating_mutex.synchronize do
+          # Emit producer.reload event before reload
+          # Users can subscribe to this event and modify event[:caller].config.kafka to change
+          # producer config. This is useful for escaping fencing loops by changing transactional.id
+          @monitor.instrument(
+            'producer.reload',
+            producer_id: id,
+            error: rd_error,
+            attempt: @transaction_fatal_error_attempts,
+            caller: self
+          )
+
+          # Clear cached state that depends on config
+          # We always clear @transactional as it might have been modified via the event
+          @transactional = nil
+
           @monitor.instrument(
             'producer.reloaded',
             producer_id: id,
             attempt: @transaction_fatal_error_attempts
           ) do
-            @client.flush(current_variant.max_wait_timeout)
-            purge
-            @client.close
-            @client = Builder.new.call(self, @config)
+            reload!
           end
         end
 

data/lib/waterdrop/producer.rb

@@ -22,12 +22,6 @@ module WaterDrop
       Rdkafka::Producer::DeliveryHandle::WaitTimeoutError
     ].freeze
 
-    # We should never reload producer on certain fatal errors as they may indicate state that
-    # cannot be recovered by simply recreating the client
-    NON_RELOADABLE_FATAL_ERRORS = %i[
-      fenced
-    ].freeze
-
     # Empty hash to save on memory allocations
     EMPTY_HASH = {}.freeze
 
@@ -35,7 +29,7 @@ module WaterDrop
     EMPTY_ARRAY = [].freeze
 
     private_constant(
-      :SUPPORTED_FLOW_ERRORS, :NON_RELOADABLE_FATAL_ERRORS, :EMPTY_HASH, :EMPTY_ARRAY
+      :SUPPORTED_FLOW_ERRORS, :EMPTY_HASH, :EMPTY_ARRAY
     )
 
     def_delegators :config
@@ -508,7 +502,7 @@ module WaterDrop
         )
 
         # Attempt to reload the producer
-        idempotent_reload_client_on_fatal_error(@idempotent_fatal_error_attempts)
+        idempotent_reload_client_on_fatal_error(@idempotent_fatal_error_attempts, e)
 
         # Wait before retrying to avoid rapid reload loops
         sleep(@config.wait_backoff_on_idempotent_fatal_error / 1_000.0)
@@ -565,5 +559,15 @@ module WaterDrop
     ensure
       @operations_in_progress.decrement
     end
+
+    # Reloads the client
+    # @note This should be used only within proper mutexes internally
+    def reload!
+      @client.flush(current_variant.max_wait_timeout)
+      purge
+      @client.close
+      @client = nil
+      @status.configured!
+    end
   end
 end

data/lib/waterdrop/version.rb

@@ -3,5 +3,5 @@
 # WaterDrop library
 module WaterDrop
   # Current WaterDrop version
-  VERSION = '2.8.12'
+  VERSION = '2.8.13'
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: waterdrop
 version: !ruby/object:Gem::Version
-  version: 2.8.12
+  version: 2.8.13
 platform: ruby
 authors:
 - Maciej Mensfeld
@@ -65,7 +65,6 @@ extensions: []
 extra_rdoc_files: []
 files:
 - ".coditsu/ci.yml"
-- ".diffend.yml"
 - ".github/CODEOWNERS"
 - ".github/FUNDING.yml"
 - ".github/ISSUE_TEMPLATE/bug_report.md"

data/.diffend.yml

@@ -1,3 +0,0 @@
-project_id: 'ee590bc9-f375-4832-ac8e-beb174b49aa5'
-shareable_id: '1325b111-b957-4fd8-bd6f-fead41e06034'
-shareable_key: 'd29c6230-8bf6-479b-83ff-98f374d3466e'