karafka-rdkafka 0.14.8 → 0.14.10

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 610db3521dfe3b63d906f2d8814756c04882cc68603a218842cc41b487124133
-  data.tar.gz: d870f56e2d7fbc1ae8d14cda5757a02d01d121740da7656deab08341904f4677
+  metadata.gz: 92d61e2b492453bf19ead6abf1c9377d5222aeba36c19c25caff5641a2c8fb1b
+  data.tar.gz: 4170931c8ced8d09813d22359c36d36da09b3f39d2429daad6a342dd3df2982c
 SHA512:
-  metadata.gz: 17bca0d972de4d24d2de1c1b39ac38bab36eff5f856dd65448215fb53a88fcc8bc0b61e856cf18802ad6649938f52b2d25fdc76b0055396821866841c8067e3a
-  data.tar.gz: 87742fba542fc3df5af16e41baa36d5a7d0c33304f37e54d6ffc50121a4dcf7173a717a0e37353b77cac8828ec707aab54056e3eba53ecfd2780dd87ef9b21ec
+  metadata.gz: ceae2da64aad6589779b60160fc28b5db6bc305c289f76bbf7f0aa034afdf88277b3aa75f4c14ec8b441250e81bca7bf7afededea22c2eb0568753cf6968c2b3
+  data.tar.gz: 61a3c8a40bca6d598782b093f6a2575fcdbe247d6a31c624f95e236015d713e3d2ace38e2fae43b622b9e4c6b37be57ef0f2d508a0a57310d3eeb79404146436

data/CHANGELOG.md

@@ -1,5 +1,12 @@
 # Rdkafka Changelog
 
+## 0.14.10 (2024-02-08)
+- [Fix] Background logger stops working after forking causing memory leaks (mensfeld).
+
+## 0.14.9 (2024-01-29)
+- [Fix] Partition cache caches invalid `nil` result for `PARTITIONS_COUNT_TTL`.
+- [Enhancement] Report `-1` instead of `nil` in case `partition_count` failure.
+
 ## 0.14.8 (2024-01-24)
 - [Enhancement] Provide support for Nix OS (alexandriainfantino)
 - [Enhancement] Skip intermediate array creation on delivery report callback execution (one per message) (mensfeld)

data/karafka-rdkafka.gemspec

@@ -3,10 +3,10 @@
 require File.expand_path('lib/rdkafka/version', __dir__)
 
 Gem::Specification.new do |gem|
-  gem.authors = ['Thijs Cadier']
+  gem.authors = ['Thijs Cadier', 'Maciej Mensfeld']
   gem.email = ["contact@karafka.io"]
   gem.description = "Modern Kafka client library for Ruby based on librdkafka"
-  gem.summary = "The rdkafka gem is a modern Kafka client library for Ruby based on librdkafka. It wraps the production-ready C client using the ffi gem and targets Kafka 1.0+ and Ruby 2.4+."
+  gem.summary = "The rdkafka gem is a modern Kafka client library for Ruby based on librdkafka. It wraps the production-ready C client using the ffi gem and targets Kafka 1.0+ and Ruby 2.7+."
   gem.license = 'MIT'
 
   gem.files = `git ls-files`.split($\)

data/lib/rdkafka/bindings.rb

@@ -147,6 +147,8 @@ module Rdkafka
                  else
                    Logger::UNKNOWN
                  end
+
+      Rdkafka::Config.ensure_log_thread
       Rdkafka::Config.log_queue << [severity, "rdkafka: #{line}"]
     end
 

data/lib/rdkafka/config.rb

@@ -15,13 +15,12 @@ module Rdkafka
     @@opaques = ObjectSpace::WeakMap.new
     # @private
     @@log_queue = Queue.new
-
-    Thread.start do
-      loop do
-        severity, msg = @@log_queue.pop
-        @@logger.add(severity, msg)
-      end
-    end
+    # @private
+    # We memoize thread on the first log flush
+    # This allows us also to restart logger thread on forks
+    @@log_thread = nil
+    # @private
+    @@log_mutex = Mutex.new
 
     # Returns the current logger, by default this is a logger to stdout.
     #
@@ -30,6 +29,24 @@ module Rdkafka
       @@logger
     end
 
+    # Makes sure that there is a thread for consuming logs
+    # We do not spawn thread immediately and we need to check if it operates to support forking
+    def self.ensure_log_thread
+      return if @@log_thread && @@log_thread.alive?
+
+      @@log_mutex.synchronize do
+        # Restart if dead (fork, crash)
+        @@log_thread = nil if @@log_thread && !@@log_thread.alive?
+
+        @@log_thread ||= Thread.start do
+          loop do
+            severity, msg = @@log_queue.pop
+            @@logger.add(severity, msg)
+          end
+        end
+      end
+    end
+
     # Returns a queue whose contents will be passed to the configured logger. Each entry
     # should follow the format [Logger::Severity, String]. The benefit over calling the
     # logger directly is that this is safe to use from trap contexts.

data/lib/rdkafka/producer.rb

@@ -40,10 +40,16 @@ module Rdkafka
           topic_metadata = ::Rdkafka::Metadata.new(inner, topic).topics&.first
         end
 
-        cache[topic] = [
-          monotonic_now,
-          topic_metadata ? topic_metadata[:partition_count] : nil
-        ]
+        partition_count = topic_metadata ? topic_metadata[:partition_count] : -1
+
+        # This approach caches the failure to fetch only for 1 second. This will make sure, that
+        # we do not cache the failure for too long but also "buys" us a bit of time in case there
+        # would be issues in the cluster so we won't overaload it with consecutive requests
+        cache[topic] = if partition_count.positive?
+                         [monotonic_now, partition_count]
+                       else
+                         [monotonic_now - PARTITIONS_COUNT_TTL + 5, partition_count]
+                       end
       end
     end
 
@@ -199,14 +205,15 @@ module Rdkafka
     # Partition count for a given topic.
     #
     # @param topic [String] The topic name.
-    # @return [Integer] partition count for a given topic
+    # @return [Integer] partition count for a given topic or `-1` if it could not be obtained.
     #
     # @note If 'allow.auto.create.topics' is set to true in the broker, the topic will be
     #   auto-created after returning nil.
     #
     # @note We cache the partition count for a given topic for given time.
     #   This prevents us in case someone uses `partition_key` from querying for the count with
-    #   each message. Instead we query once every 30 seconds at most
+    #   each message. Instead we query once every 30 seconds at most if we have a valid partition
+    #   count or every 5 seconds in case we were not able to obtain number of partitions
     def partition_count(topic)
       closed_producer_check(__method__)
 
@@ -256,7 +263,7 @@ module Rdkafka
       if partition_key
         partition_count = partition_count(topic)
         # If the topic is not present, set to -1
-        partition = Rdkafka::Bindings.partitioner(partition_key, partition_count, @partitioner_name) if partition_count
+        partition = Rdkafka::Bindings.partitioner(partition_key, partition_count, @partitioner_name) if partition_count.positive?
       end
 
       # If partition is nil, use -1 to let librdafka set the partition randomly or

data/lib/rdkafka/version.rb

@@ -1,7 +1,7 @@
 # frozen_string_literal: true
 
 module Rdkafka
-  VERSION = "0.14.8"
+  VERSION = "0.14.10"
   LIBRDKAFKA_VERSION = "2.3.0"
   LIBRDKAFKA_SOURCE_SHA256 = "2d49c35c77eeb3d42fa61c43757fcbb6a206daa560247154e60642bcdcc14d12"
 end

data/spec/rdkafka/config_spec.rb

@@ -22,6 +22,7 @@ describe Rdkafka::Config do
     it "supports logging queue" do
       log = StringIO.new
       Rdkafka::Config.logger = Logger.new(log)
+      Rdkafka::Config.ensure_log_thread
 
       Rdkafka::Config.log_queue << [Logger::FATAL, "I love testing"]
       20.times do
@@ -31,6 +32,25 @@ describe Rdkafka::Config do
 
       expect(log.string).to include "FATAL -- : I love testing"
     end
+
+    it "expect to start new logger thread after fork and work" do
+      reader, writer = IO.pipe
+
+      pid = fork do
+        $stdout.reopen(writer)
+        Rdkafka::Config.logger = Logger.new($stdout)
+        reader.close
+        producer = rdkafka_producer_config(debug: 'all').producer
+        producer.close
+        writer.close
+        sleep(1)
+      end
+
+      writer.close
+      Process.wait(pid)
+      output = reader.read
+      expect(output.split("\n").size).to be >= 20
+    end
   end
 
   context "statistics callback" do

data/spec/rdkafka/consumer_spec.rb

@@ -211,6 +211,11 @@ describe Rdkafka::Consumer do
 
         # 7. ensure same message is read again
         message2 = consumer.poll(timeout)
+
+        # This is needed because `enable.auto.offset.store` is true but when running in CI that
+        # is overloaded, offset store lags
+        sleep(1)
+
         consumer.commit
         expect(message1.offset).to eq message2.offset
         expect(message1.payload).to eq message2.payload

metadata

@@ -1,10 +1,11 @@
 --- !ruby/object:Gem::Specification
 name: karafka-rdkafka
 version: !ruby/object:Gem::Version
-  version: 0.14.8
+  version: 0.14.10
 platform: ruby
 authors:
 - Thijs Cadier
+- Maciej Mensfeld
 autorequire: 
 bindir: bin
 cert_chain:
@@ -35,7 +36,7 @@ cert_chain:
   AnG1dJU+yL2BK7vaVytLTstJME5mepSZ46qqIJXMuWob/YPDmVaBF39TDSG9e34s
   msG3BiCqgOgHAnL23+CN3Rt8MsuRfEtoTKpJVcCfoEoNHOkc
   -----END CERTIFICATE-----
-date: 2024-01-24 00:00:00.000000000 Z
+date: 2024-02-08 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ffi
@@ -282,7 +283,7 @@ signing_key:
 specification_version: 4
 summary: The rdkafka gem is a modern Kafka client library for Ruby based on librdkafka.
   It wraps the production-ready C client using the ffi gem and targets Kafka 1.0+
-  and Ruby 2.4+.
+  and Ruby 2.7+.
 test_files:
 - spec/rdkafka/abstract_handle_spec.rb
 - spec/rdkafka/admin/create_acl_handle_spec.rb