karafka 2.1.10 → 2.1.11

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 78db49b9cd20426b92ad1a3417b9bc2e79e9efad7f7d11e42873ffda75b850c2
-  data.tar.gz: 40b2a338c11ad9d6d07fdb0b82f422fe8f64417805db85a9bd111f9bce29039f
+  metadata.gz: 8d1e3000ced82b96afc7c4588d20170e0908d3b41305eead2e5e0a850c9a70c4
+  data.tar.gz: 1b16fc0a06f1bfd9bdb1207dd1faac04d3a7f956ccb42edd45ada76fed6dcf09
 SHA512:
-  metadata.gz: a522abfbe63ebde9255032b5c6760016a014b8952f8c3f593c754479921408451ac0aeca632ba32d1fcefac78acd97a110e777949f6e70587d454febf5ad34f5
-  data.tar.gz: d12c97a4829bb0278041ce20e38f21c4894c77a7c9d4b3e4c1c222c502632fe03b0cce3d224f4b8c28ec3743c9c51e41c98a53310761687e92ce587445eae300
+  metadata.gz: 7e018450fae0ad666bff80ca9875750303946746c20e527200ee569a1a182c3d81bb71b64a0b2bd94b4a35c8cf204326b26752b1d5c0425a1f544ff9b4572323
+  data.tar.gz: 6025b3583179592313540231dd9d00354c30d8af3d49f482fc6c7dbd96359dbc7bbccee8cfa4cf1d14a2f5dfb3a9999b141317c782ec33af7503c4a8e26c616c

data/CHANGELOG.md

@@ -1,5 +1,9 @@
 # Karafka framework changelog
 
+## 2.1.11 (2023-08-23)
+- [Enhancement] Expand the error handling for offset related queries with timeout error retries.
+- [Enhancement] Allow for connection proxy timeouts configuration.
+
 ## 2.1.10 (2023-08-21)
 - [Enhancement] Introduce `connection.client.rebalance_callback` event for instrumentation of rebalances.
 - [Refactor] Introduce low level commands proxy to handle deviation in how we want to run certain commands and how rdkafka-ruby runs that by design.

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    karafka (2.1.10)
+    karafka (2.1.11)
       karafka-core (>= 2.1.1, < 2.2.0)
       thor (>= 0.20)
       waterdrop (>= 2.6.6, < 3.0.0)
@@ -33,7 +33,7 @@ GEM
     karafka-core (2.1.1)
       concurrent-ruby (>= 1.1)
       karafka-rdkafka (>= 0.13.1, < 0.14.0)
-    karafka-rdkafka (0.13.3)
+    karafka-rdkafka (0.13.4)
       ffi (~> 1.15)
       mini_portile2 (~> 2.6)
       rake (> 12)

data/config/locales/errors.yml

@@ -26,6 +26,12 @@ en:
       internal.process_format: needs to be present
       internal.routing.builder_format: needs to be present
       internal.routing.subscription_groups_builder_format: needs to be present
+      internal.connection.proxy.query_watermark_offsets.timeout_format: needs to be an integer bigger than 0
+      internal.connection.proxy.query_watermark_offsets.max_attempts_format: needs to be an integer bigger than 0
+      internal.connection.proxy.query_watermark_offsets.wait_time_format: needs to be an integer bigger than 0
+      internal.connection.proxy.offsets_for_times.timeout_format: needs to be an integer bigger than 0
+      internal.connection.proxy.offsets_for_times.max_attempts_format: needs to be an integer bigger than 0
+      internal.connection.proxy.offsets_for_times.wait_time_format: needs to be an integer bigger than 0
       key_must_be_a_symbol: All keys under the kafka settings scope need to be symbols
       max_timeout_vs_pause_max_timeout: pause_timeout must be less or equal to pause_max_timeout
       shutdown_timeout_vs_max_wait_time: shutdown_timeout must be more than max_wait_time

data/lib/karafka/connection/proxy.rb

@@ -10,20 +10,13 @@ module Karafka
     # do still want to be able to alter some functionalities. This wrapper helps us do it when
     # it would be needed
     class Proxy < SimpleDelegator
-      # Timeout on the watermark query
-      WATERMARK_REQUEST_TIMEOUT = 5_000
+      # Errors on which we want to retry
+      RETRYABLE_ERRORS = %i[
+        all_brokers_down
+        timed_out
+      ].freeze
 
-      # Timeout on the TPL request query
-      TPL_REQUEST_TIMEOUT = 5_000
-
-      # How many attempts we want to take for something that would end up with all_brokers_down
-      BROKERS_DOWN_MAX_ATTEMPTS = 3
-
-      # How long should we wait in between all_brokers_down
-      BROKERS_DOWN_BACKOFF_TIME = 1
-
-      private_constant :WATERMARK_REQUEST_TIMEOUT, :BROKERS_DOWN_MAX_ATTEMPTS,
-                       :BROKERS_DOWN_BACKOFF_TIME, :TPL_REQUEST_TIMEOUT
+      private_constant :RETRYABLE_ERRORS
 
       attr_accessor :wrapped
 
@@ -36,6 +29,7 @@ module Karafka
         # wrap an already wrapped object with another proxy level. Simplifies passing consumers
         # and makes it safe to wrap without type checking
         @wrapped = obj.is_a?(self.class) ? obj.wrapped : obj
+        @config = ::Karafka::App.config.internal.connection.proxy
       end
 
       # Proxies the `#query_watermark_offsets` with extra recovery from timeout problems.
@@ -46,12 +40,14 @@ module Karafka
       # @param partition [Partition]
       # @return [Array<Integer, Integer>] watermark offsets
       def query_watermark_offsets(topic, partition)
-        with_brokers_down_retry do
-          @wrapped.query_watermark_offsets(
-            topic,
-            partition,
-            WATERMARK_REQUEST_TIMEOUT
-          )
+        l_config = @config.query_watermark_offsets
+
+        with_broker_errors_retry(
+          # required to be in seconds, not ms
+          wait_time: l_config.wait_time / 1_000.to_f,
+          max_attempts: l_config.max_attempts
+        ) do
+          @wrapped.query_watermark_offsets(topic, partition, l_config.timeout)
         end
       end
 
@@ -61,8 +57,14 @@ module Karafka
       # @param tpl [Rdkafka::Consumer::TopicPartitionList] tpl to get time offsets
       # @return [Rdkafka::Consumer::TopicPartitionList] tpl with time offsets
       def offsets_for_times(tpl)
-        with_brokers_down_retry do
-          @wrapped.offsets_for_times(tpl, TPL_REQUEST_TIMEOUT)
+        l_config = @config.offsets_for_times
+
+        with_broker_errors_retry(
+          # required to be in seconds, not ms
+          wait_time: l_config.wait_time / 1_000.to_f,
+          max_attempts: l_config.max_attempts
+        ) do
+          @wrapped.offsets_for_times(tpl, l_config.timeout)
         end
       end
 
@@ -71,16 +73,20 @@ module Karafka
       # Runs expected block of code with few retries on all_brokers_down
       # librdkafka can return `all_brokers_down` for scenarios when broker is overloaded or not
       # reachable due to latency.
-      def with_brokers_down_retry
+      # @param max_attempts [Integer] how many attempts (not retries) should we take before failing
+      #   completely.
+      # @param wait_time [Integer, Float] how many seconds should we wait. It uses `#sleep` of Ruby
+      #   so it needs time in seconds.
+      def with_broker_errors_retry(max_attempts:, wait_time: 1)
         attempt ||= 0
         attempt += 1
 
         yield
       rescue Rdkafka::RdkafkaError => e
-        raise if e.code != :all_brokers_down
+        raise unless RETRYABLE_ERRORS.include?(e.code)
 
-        if attempt <= BROKERS_DOWN_MAX_ATTEMPTS
-          sleep(BROKERS_DOWN_BACKOFF_TIME)
+        if attempt <= max_attempts
+          sleep(wait_time)
 
           retry
         end

data/lib/karafka/contracts/config.rb

@@ -39,6 +39,22 @@ module Karafka
         required(:status) { |val| !val.nil? }
         required(:process) { |val| !val.nil? }
 
+        nested(:connection) do
+          nested(:proxy) do
+            nested(:query_watermark_offsets) do
+              required(:timeout) { |val| val.is_a?(Integer) && val.positive? }
+              required(:max_attempts) { |val| val.is_a?(Integer) && val.positive? }
+              required(:wait_time) { |val| val.is_a?(Integer) && val.positive? }
+            end
+
+            nested(:offsets_for_times) do
+              required(:timeout) { |val| val.is_a?(Integer) && val.positive? }
+              required(:max_attempts) { |val| val.is_a?(Integer) && val.positive? }
+              required(:wait_time) { |val| val.is_a?(Integer) && val.positive? }
+            end
+          end
+        end
+
         nested(:routing) do
           required(:builder) { |val| !val.nil? }
           required(:subscription_groups_builder) { |val| !val.nil? }

data/lib/karafka/setup/config.rb

@@ -126,6 +126,32 @@ module Karafka
           setting :activity_manager, default: Routing::ActivityManager.new
         end
 
+        # Namespace for internal connection related settings
+        setting :connection do
+          # Settings that are altered by our client proxy layer
+          setting :proxy do
+            # Watermark offsets request settings
+            setting :query_watermark_offsets do
+              # timeout for this request. For busy or remote clusters, this should be high enough
+              setting :timeout, default: 5_000
+              # How many times should we try to run this call before raising an error
+              setting :max_attempts, default: 3
+              # How long should we wait before next attempt in case of a failure
+              setting :wait_time, default: 1_000
+            end
+
+            # Offsets for times request settings
+            setting :offsets_for_times do
+              # timeout for this request. For busy or remote clusters, this should be high enough
+              setting :timeout, default: 5_000
+              # How many times should we try to run this call before raising an error
+              setting :max_attempts, default: 3
+              # How long should we wait before next attempt in case of a failure
+              setting :wait_time, default: 1_000
+            end
+          end
+        end
+
         setting :processing do
           # option scheduler [Object] scheduler we will be using
           setting :scheduler, default: Processing::Scheduler.new

data/lib/karafka/version.rb

@@ -3,5 +3,5 @@
 # Main module namespace
 module Karafka
   # Current Karafka version
-  VERSION = '2.1.10'
+  VERSION = '2.1.11'
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: karafka
 version: !ruby/object:Gem::Version
-  version: 2.1.10
+  version: 2.1.11
 platform: ruby
 authors:
 - Maciej Mensfeld
@@ -35,7 +35,7 @@ cert_chain:
   AnG1dJU+yL2BK7vaVytLTstJME5mepSZ46qqIJXMuWob/YPDmVaBF39TDSG9e34s
   msG3BiCqgOgHAnL23+CN3Rt8MsuRfEtoTKpJVcCfoEoNHOkc
   -----END CERTIFICATE-----
-date: 2023-08-21 00:00:00.000000000 Z
+date: 2023-08-23 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: karafka-core