karafka 1.4.9 → 2.0.0.alpha1

data/lib/karafka/setup/config.rb

@@ -17,14 +17,33 @@ module Karafka
       # Contract for checking the config provided by the user
       CONTRACT = Karafka::Contracts::Config.new.freeze
 
-      private_constant :CONTRACT
+      # Defaults for kafka settings, that will be overwritten only if not present already
+      KAFKA_DEFAULTS = {
+        'client.id' => 'karafka'
+      }.freeze
+
+      private_constant :CONTRACT, :KAFKA_DEFAULTS
 
       # Available settings
+
+      # Namespace for Pro version related license management. If you use LGPL, no need to worry
+      #   about any of this
+      setting :license do
+        # option token [String, false] - license token issued when you acquire a Pro license
+        # Leave false if using the LGPL version and all is going to work just fine :)
+        #
+        # @note By using the commercial components, you accept the LICENSE-COMM commercial license
+        #   terms and conditions
+        setting :token, default: false
+        # option entity [String] for whom we did issue the license
+        setting :entity, default: ''
+        # option expires_on [Date] date when the license expires
+        setting :expires_on, default: Date.parse('2100-01-01')
+      end
+
       # option client_id [String] kafka client_id - used to provide
       #   default Kafka groups namespaces and identify that app in kafka
-      setting :client_id
-      # What backend do we want to use to process messages
-      setting :backend, default: :inline
+      setting :client_id, default: 'karafka'
       # option logger [Instance] logger that we want to use
       setting :logger, default: ::Karafka::Instrumentation::Logger.new
       # option monitor [Instance] monitor that we will to use (defaults to Karafka::Monitor)
@@ -33,147 +52,36 @@ module Karafka
       # or they need to maintain their own internal consumer group naming conventions, they
       # can easily do it, replacing the default client_id + consumer name pattern concept
       setting :consumer_mapper, default: Routing::ConsumerMapper.new
-      # Mapper used to remap names of topics, so we can have a clean internal topic naming
-      # despite using any Kafka provider that uses namespacing, etc
-      # It needs to implement two methods:
-      #   - #incoming - for remapping from the incoming message to our internal format
-      #   - #outgoing - for remapping from internal topic name into outgoing message
-      setting :topic_mapper, default: Routing::TopicMapper.new
-      # Default serializer for converting whatever we want to send to kafka to json
-      setting :serializer, default: Karafka::Serialization::Json::Serializer.new
+      # option [Boolean] should we reload consumers with each incoming batch thus effectively
+      # supporting code reload (if someone reloads code) or should we keep the persistence
+      setting :consumer_persistence, default: true
       # Default deserializer for converting incoming data into ruby objects
       setting :deserializer, default: Karafka::Serialization::Json::Deserializer.new
-      # If batch_fetching is true, we will fetch kafka messages in batches instead of 1 by 1
-      # @note Fetching does not equal consuming, see batch_consuming description for details
-      setting :batch_fetching, default: true
-      # If batch_consuming is true, we will have access to #params_batch instead of #params.
-      # #params_batch will contain params received from Kafka (may be more than 1) so we can
-      # process them in batches
-      setting :batch_consuming, default: false
-      # option shutdown_timeout [Integer, nil] the number of seconds after which Karafka no
-      #   longer wait for the consumers to stop gracefully but instead we force terminate
+      # option [Boolean] should we leave offset management to the user
+      setting :manual_offset_management, default: false
+      # options max_messages [Integer] how many messages do we want to fetch from Kafka in one go
+      setting :max_messages, default: 100_000
+      # option [Integer] number of milliseconds we can wait while fetching data
+      setting :max_wait_time, default: 10_000
+      # option shutdown_timeout [Integer] the number of milliseconds after which Karafka no
+      #   longer waits for the consumers to stop gracefully but instead we force terminate
       #   everything.
-      setting :shutdown_timeout, default: 60
-
-      # option kafka [Hash] - optional - kafka configuration options
-      setting :kafka do
-        # Array with at least one host
-        setting :seed_brokers, default: %w[kafka://127.0.0.1:9092]
-        # option session_timeout [Integer] the number of seconds after which, if a client
-        #   hasn't contacted the Kafka cluster, it will be kicked out of the group.
-        setting :session_timeout, default: 30
-        # Time that a given partition will be paused from fetching messages, when message
-        # consumption fails. It allows us to process other partitions, while the error is being
-        # resolved and also "slows" things down, so it prevents from "eating" up all messages and
-        # consuming them with failed code. Use `nil` if you want to pause forever and never retry.
-        setting :pause_timeout, default: 10
-        # option pause_max_timeout [Integer, nil] the maximum number of seconds to pause for,
-        #   or `nil` if no maximum should be enforced.
-        setting :pause_max_timeout, default: nil
-        # option pause_exponential_backoff [Boolean] whether to enable exponential backoff
-        setting :pause_exponential_backoff, default: false
-        # option offset_commit_interval [Integer] the interval between offset commits,
-        #   in seconds.
-        setting :offset_commit_interval, default: 10
-        # option offset_commit_threshold [Integer] the number of messages that can be
-        #   processed before their offsets are committed. If zero, offset commits are
-        #   not triggered by message consumption.
-        setting :offset_commit_threshold, default: 0
-        # option heartbeat_interval [Integer] the interval between heartbeats; must be less
-        #   than the session window.
-        setting :heartbeat_interval, default: 10
-        # option offset_retention_time [Integer] The length of the retention window, known as
-        #   offset retention time
-        setting :offset_retention_time, default: nil
-        # option fetcher_max_queue_size [Integer] max number of items in the fetch queue that
-        #   are stored for further processing. Note, that each item in the queue represents a
-        #   response from a single broker
-        setting :fetcher_max_queue_size, default: 10
-        # option assignment_strategy [Object] a strategy determining the assignment of
-        #   partitions to the consumers.
-        setting :assignment_strategy, default: Karafka::AssignmentStrategies::RoundRobin.new
-        # option max_bytes_per_partition [Integer] the maximum amount of data fetched
-        #   from a single partition at a time.
-        setting :max_bytes_per_partition, default: 1_048_576
-        #  whether to consume messages starting at the beginning or to just consume new messages
-        setting :start_from_beginning, default: true
-        # option resolve_seed_brokers [Boolean] whether to resolve each hostname of the seed
-        # brokers
-        setting :resolve_seed_brokers, default: false
-        # option min_bytes [Integer] the minimum number of bytes to read before
-        #   returning messages from the server; if `max_wait_time` is reached, this
-        #   is ignored.
-        setting :min_bytes, default: 1
-        # option max_bytes [Integer] the maximum number of bytes to read before returning messages
-        #   from each broker.
-        setting :max_bytes, default: 10_485_760
-        # option max_wait_time [Integer, Float] max_wait_time is the maximum number of seconds to
-        #   wait before returning data from a single message fetch. By setting this high you also
-        #   increase the fetching throughput - and by setting it low you set a bound on latency.
-        #   This configuration overrides `min_bytes`, so you'll _always_ get data back within the
-        #   time specified. The default value is one second. If you want to have at most five
-        #   seconds of latency, set `max_wait_time` to 5. You should make sure
-        #   max_wait_time * num brokers + heartbeat_interval is less than session_timeout.
-        setting :max_wait_time, default: 1
-        # option automatically_mark_as_consumed [Boolean] should we automatically mark received
-        # messages as consumed (processed) after non-error consumption
-        setting :automatically_mark_as_consumed, default: true
-        # option reconnect_timeout [Integer] How long should we wait before trying to reconnect to
-        # Kafka cluster that went down (in seconds)
-        setting :reconnect_timeout, default: 5
-        # option connect_timeout [Integer] Sets the number of seconds to wait while connecting to
-        # a broker for the first time. When ruby-kafka initializes, it needs to connect to at
-        # least one host.
-        setting :connect_timeout, default: 10
-        # option socket_timeout [Integer] Sets the number of seconds to wait when reading from or
-        # writing to a socket connection to a broker. After this timeout expires the connection
-        # will be killed. Note that some Kafka operations are by definition long-running, such as
-        # waiting for new messages to arrive in a partition, so don't set this value too low
-        setting :socket_timeout, default: 30
-        # option partitioner [Object, nil] the partitioner that should be used by the client
-        setting :partitioner, default: nil
-
-        # SSL authentication related settings
-        # option ca_cert [String, nil] SSL CA certificate
-        setting :ssl_ca_cert, default: nil
-        # option ssl_ca_cert_file_path [String, nil] SSL CA certificate file path
-        setting :ssl_ca_cert_file_path, default: nil
-        # option ssl_ca_certs_from_system [Boolean] Use the CA certs from your system's default
-        #   certificate store
-        setting :ssl_ca_certs_from_system, default: false
-        # option ssl_verify_hostname [Boolean] Verify the hostname for client certs
-        setting :ssl_verify_hostname, default: true
-        # option ssl_client_cert [String, nil] SSL client certificate
-        setting :ssl_client_cert, default: nil
-        # option ssl_client_cert_key [String, nil] SSL client certificate password
-        setting :ssl_client_cert_key, default: nil
-        # option sasl_gssapi_principal [String, nil] sasl principal
-        setting :sasl_gssapi_principal, default: nil
-        # option sasl_gssapi_keytab [String, nil] sasl keytab
-        setting :sasl_gssapi_keytab, default: nil
-        # option sasl_plain_authzid [String] The authorization identity to use
-        setting :sasl_plain_authzid, default: ''
-        # option sasl_plain_username [String, nil] The username used to authenticate
-        setting :sasl_plain_username, default: nil
-        # option sasl_plain_password [String, nil] The password used to authenticate
-        setting :sasl_plain_password, default: nil
-        # option sasl_scram_username [String, nil] The username used to authenticate
-        setting :sasl_scram_username, default: nil
-        # option sasl_scram_password [String, nil] The password used to authenticate
-        setting :sasl_scram_password, default: nil
-        # option sasl_scram_mechanism [String, nil] Scram mechanism, either 'sha256' or 'sha512'
-        setting :sasl_scram_mechanism, default: nil
-        # option sasl_over_ssl [Boolean] whether to enforce SSL with SASL
-        setting :sasl_over_ssl, default: true
-        # option ssl_client_cert_chain [String, nil] client cert chain or nil if not used
-        setting :ssl_client_cert_chain, default: nil
-        # option ssl_client_cert_key_password [String, nil] the password required to read
-        #   the ssl_client_cert_key
-        setting :ssl_client_cert_key_password, default: nil
-        # @param sasl_oauth_token_provider [Object, nil] OAuthBearer Token Provider instance that
-        #   implements method token.
-        setting :sasl_oauth_token_provider, default: nil
-      end
+      setting :shutdown_timeout, default: 60_000
+      # option [Integer] number of threads in which we want to do parallel processing
+      setting :concurrency, default: 5
+      # option [Integer] how long should we wait upon processing error
+      setting :pause_timeout, default: 1_000
+      # option [Integer] what is the max timeout in case of an exponential backoff
+      setting :pause_max_timeout, default: 30_000
+      # option [Boolean] should we use exponential backoff
+      setting :pause_with_exponential_backoff, default: true
+      # option [::WaterDrop::Producer, nil]
+      # Unless configured, will be created once Karafka is configured based on user Karafka setup
+      setting :producer, default: nil
+
+      # rdkafka default options
+      # @see https://github.com/edenhill/librdkafka/blob/master/CONFIGURATION.md
+      setting :kafka, default: {}
 
       # Namespace for internal settings that should not be modified
       # It's a temporary step to "declassify" several things internally before we move to a
@@ -187,29 +95,37 @@ module Karafka
         # @note In the future, we need to have a single process representation for all the karafka
         #   instances
         setting :process, default: Process.new
-        # option fetcher [Karafka::Fetcher] fetcher instance
-        setting :fetcher, default: Fetcher.new
-        # option configurators [Array<Object>] all configurators that we want to run after
-        #   the setup
-        setting :configurators, default: [Configurators::WaterDrop.new]
+        # option subscription_groups_builder [Routing::SubscriptionGroupsBuilder] subscription
+        #   group builder
+        setting :subscription_groups_builder, default: Routing::SubscriptionGroupsBuilder.new
       end
 
       class << self
         # Configuring method
-        # @yield Runs a block of code providing a config singleton instance to it
-        # @yieldparam [Karafka::Setup::Config] Karafka config instance
-        def setup
-          configure { |config| yield(config) }
+        # @param block [Proc] block we want to execute with the config instance
+        def setup(&block)
+          configure(&block)
+          merge_kafka_defaults!(config)
+          validate!
+
+          # Check the license presence (if needed) and
+          Licenser.new.verify(config.license)
+
+          configure_components
         end
 
-        # Everything that should be initialized after the setup
-        # Components are in karafka/config directory and are all loaded one by one
-        # If you want to configure a next component, please add a proper file to config dir
-        def setup_components
-          config
-            .internal
-            .configurators
-            .each { |configurator| configurator.call(config) }
+        private
+
+        # Propagates the kafka setting defaults unless they are already present
+        # This makes it easier to set some values that users usually don't change but still allows
+        # them to overwrite the whole hash if they want to
+        # @param config [Dry::Configurable::Config] dry config of this producer
+        def merge_kafka_defaults!(config)
+          KAFKA_DEFAULTS.each do |key, value|
+            next if config.kafka.key?(key)
+
+            config.kafka[key] = value
+          end
         end
 
         # Validate config based on the config contract
@@ -223,6 +139,17 @@ module Karafka
 
           raise Errors::InvalidConfigurationError, validation_result.errors.to_h
         end
+
+        # Sets up all the components that are based on the user configuration
+        # @note At the moment it is only WaterDrop
+        def configure_components
+          config.producer ||= ::WaterDrop::Producer.new do |producer_config|
+            # In some cases WaterDrop updates the config and we don't want our consumer config to
+            # be polluted by those updates, that's why we copy
+            producer_config.kafka = config.kafka.dup
+            producer_config.logger = config.logger
+          end
+        end
       end
     end
   end

data/lib/karafka/status.rb

@@ -3,16 +3,21 @@
 module Karafka
   # App status monitor
   class Status
-    # Available states and their transitions
+    # Available states and their transitions.
     STATES = {
       initializing: :initialize!,
-      initialized: :initialized!,
       running: :run!,
-      stopping: :stop!
+      stopping: :stop!,
+      stopped: :stopped!
     }.freeze
 
     private_constant :STATES
 
+    # By default we are in the initializing state
+    def initialize
+      initialize!
+    end
+
     STATES.each do |state, transition|
       define_method :"#{state}?" do
         @status == state
@@ -20,6 +25,11 @@ module Karafka
 
       define_method transition do
         @status = state
+
+        # Skip on creation (initializing)
+        # We skip as during this state we do not have yet a monitor
+        return if initializing?
+
         # Trap context disallows to run certain things that we instrument
         # so the state changes are executed from a separate thread
         Thread.new { Karafka.monitor.instrument("app.#{state}") }.join

data/lib/karafka/templates/example_consumer.rb.erb

@@ -0,0 +1,16 @@
+# frozen_string_literal: true
+
+# Example consumer that prints messages payloads
+class ExampleConsumer < ApplicationConsumer
+  def consume
+    messages.each { |message| puts message.payload }
+  end
+
+  # Run anything upon partition being revoked
+  # def on_revoked
+  # end
+
+  # Define here any teardown things you want when Karafka server stops
+  # def on_shutdown
+  # end
+end

data/lib/karafka/templates/karafka.rb.erb

@@ -1,19 +1,7 @@
 # frozen_string_literal: true
 
 <% if rails? -%>
-ENV['RAILS_ENV'] ||= 'development'
-ENV['KARAFKA_ENV'] = ENV['RAILS_ENV']
 require ::File.expand_path('../config/environment', __FILE__)
-Rails.application.eager_load!
-
-# This lines will make Karafka print to stdout like puma or unicorn
-if Rails.env.development?
-  Rails.logger.extend(
-    ActiveSupport::Logger.broadcast(
-      ActiveSupport::Logger.new($stdout)
-    )
-  )
-end
 <% else -%>
 # This file is auto-generated during the install process.
 # If by any chance you've wanted a setup for Rails app, either run the `karafka:install`
@@ -31,9 +19,7 @@ APP_LOADER.enable_reloading
 %w[
   lib
   app/consumers
-  app/responders
-  app/workers
-].each(&APP_LOADER.method(:push_dir))
+].each { |dir| APP_LOADER.push_dir(dir) }
 
 APP_LOADER.setup
 APP_LOADER.eager_load
@@ -41,10 +27,12 @@ APP_LOADER.eager_load
 
 class KarafkaApp < Karafka::App
   setup do |config|
-    config.kafka.seed_brokers = %w[kafka://127.0.0.1:9092]
+    config.kafka = { 'bootstrap.servers' => '127.0.0.1:9092' }
     config.client_id = 'example_app'
 <% if rails? -%>
-    config.logger = Rails.logger
+    # Recreate consumers with each batch. This will allow Rails code reload to work in the
+    # development mode. Otherwise Karafka process would not be aware of code changes
+    config.consumer_persistence = !Rails.env.development?
 <% end -%>
   end
 
@@ -52,41 +40,17 @@ class KarafkaApp < Karafka::App
   # interested in logging events for certain environments. Since instrumentation
   # notifications add extra boilerplate, if you want to achieve max performance,
   # listen to only what you really need for given environment.
-  Karafka.monitor.subscribe(WaterDrop::Instrumentation::StdoutListener.new)
   Karafka.monitor.subscribe(Karafka::Instrumentation::StdoutListener.new)
   # Karafka.monitor.subscribe(Karafka::Instrumentation::ProctitleListener.new)
 
-  # Uncomment that in order to achieve code reload in development mode
-  # Be aware, that this might have some side-effects. Please refer to the wiki
-  # for more details on benefits and downsides of the code reload in the
-  # development mode
-  #
-  # Karafka.monitor.subscribe(
-  #   Karafka::CodeReloader.new(
-  #     <%= rails? ? '*Rails.application.reloaders' : 'APP_LOADER' %>
-  #   )
-  # )
-
-  consumer_groups.draw do
-    # topic :example do
-    #   consumer ExampleConsumer
-    # end
-
-    # consumer_group :bigger_group do
-    #   topic :test do
-    #     consumer TestConsumer
-    #   end
-    #
-    #   topic :test2 do
-    #     consumer Test2Consumer
-    #   end
-    # end
+  routes.draw do
+<% if rails? -%>
+    # Uncomment this if you use Karafka with ActiveJob
+    # You ned to define the topic per each queue name you use
+    # active_job_topic :default
+<% end -%>
+    topic :example do
+      consumer ExampleConsumer
+    end
   end
 end
-
-Karafka.monitor.subscribe('app.initialized') do
-  # Put here all the things you want to do after the Karafka framework
-  # initialization
-end
-
-KarafkaApp.boot!

data/lib/karafka/time_trackers/base.rb

@@ -0,0 +1,19 @@
+# frozen_string_literal: true
+
+module Karafka
+  # Time trackers module.
+  #
+  # Time trackers are used to track time in context of having a time poll (amount of time
+  # available for processing) or a pausing engine (pause for a time period).
+  module TimeTrackers
+    # Base class for all the time-trackers.
+    class Base
+      private
+
+      # @return [Float] current time in milliseconds
+      def now
+        ::Process.clock_gettime(::Process::CLOCK_MONOTONIC) * 1000
+      end
+    end
+  end
+end

data/lib/karafka/time_trackers/pause.rb

@@ -0,0 +1,84 @@
+# frozen_string_literal: true
+
+module Karafka
+  module TimeTrackers
+    # Handles Kafka topic partition pausing and resuming with exponential back-offs.
+    class Pause < Base
+      attr_reader :count
+
+      # @param timeout [Integer] how long should we wait when anything went wrong (in ms)
+      # @param max_timeout [Integer, nil] if exponential is on, what is the max value we can reach
+      #   exponentially on which we will stay
+      # @param exponential_backoff [Boolean] should we wait exponentially or with the same
+      #   timeout value
+      # @return [Karafka::TimeTrackers::Pause]
+      # @example
+      #   pause = Karafka::TimeTrackers::Pause.new(timeout: 1000)
+      #   pause.expired? #=> true
+      #   pause.paused? #=> false
+      #   pause.pause
+      #   sleep(1.1)
+      #   pause.paused? #=> true
+      #   pause.expired? #=> true
+      #   pause.count #=> 1
+      #   pause.pause
+      #   pause.count #=> 1
+      #   pause.paused? #=> true
+      #   pause.expired? #=> false
+      #   pause.resume
+      #   pause.count #=> 2
+      #   pause.paused? #=> false
+      #   pause.reset
+      #   pause.count #=> 0
+      def initialize(timeout:, max_timeout:, exponential_backoff:)
+        @started_at = nil
+        @count = 0
+        @timeout = timeout
+        @max_timeout = max_timeout
+        @exponential_backoff = exponential_backoff
+        super()
+      end
+
+      # Pauses the processing from now till the end of the interval (backoff or non-backoff)
+      # and records the count.
+      def pause
+        @started_at = now
+        @ends_at = @started_at + backoff_interval
+        @count += 1
+      end
+
+      # Marks the pause as resumed.
+      def resume
+        @started_at = nil
+        @ends_at = nil
+      end
+
+      # @return [Boolean] are we paused from processing
+      def paused?
+        !@started_at.nil?
+      end
+
+      # @return [Boolean] did the pause expire
+      def expired?
+        @ends_at ? now >= @ends_at : true
+      end
+
+      # Resets the pause counter.
+      def reset
+        @count = 0
+      end
+
+      private
+
+      # Computers the exponential backoff
+      # @return [Integer] backoff in milliseconds
+      def backoff_interval
+        backoff_factor = @exponential_backoff ? 2**@count : 1
+
+        timeout = backoff_factor * @timeout
+
+        @max_timeout && timeout > @max_timeout ? @max_timeout : timeout
+      end
+    end
+  end
+end

data/lib/karafka/time_trackers/poll.rb

@@ -0,0 +1,65 @@
+# frozen_string_literal: true
+
+module Karafka
+  module TimeTrackers
+    # Object used to keep track of time we've used running certain operations.
+    #
+    # @example Keep track of sleeping and stop after 3 seconds of 0.1 sleep intervals
+    #   time_poll = Poll.new(3000)
+    #   time_poll.start
+    #
+    #   until time_poll.exceeded?
+    #     time_poll.start
+    #     puts "I have #{time_poll.remaining.to_i}ms remaining to sleep..."
+    #     sleep(0.1)
+    #     time_poll.checkpoint
+    #   end
+    class Poll < Base
+      attr_reader :remaining, :attempts
+
+      # @param total_time [Integer] amount of milliseconds before we exceed the given time limit
+      # @return [TimeTracker] time poll instance
+      def initialize(total_time)
+        @remaining = total_time
+        @attempts = 0
+        super()
+      end
+
+      # @return [Boolean] did we exceed the time limit
+      def exceeded?
+        @remaining <= 0
+      end
+
+      # Starts time tracking.
+      def start
+        @attempts += 1
+        @started_at = now
+      end
+
+      # Stops time tracking of a given piece of code and updates the remaining time.
+      def checkpoint
+        @remaining -= (now - @started_at)
+      end
+
+      # @return [Boolean] If anything went wrong, can we retry after a backoff period or not
+      #   (do we have enough time)
+      def retryable?
+        remaining > backoff_interval
+      end
+
+      # Sleeps for amount of time matching attempt, so we sleep more with each attempt in case of
+      #   a retry.
+      def backoff
+        # Sleep requires seconds not ms
+        sleep(backoff_interval / 1_000.0)
+      end
+
+      private
+
+      # @return [Integer] milliseconds of the backoff time
+      def backoff_interval
+        100 * attempts
+      end
+    end
+  end
+end

data/lib/karafka/version.rb

@@ -3,5 +3,5 @@
 # Main module namespace
 module Karafka
   # Current Karafka version
-  VERSION = '1.4.9'
+  VERSION = '2.0.0.alpha1'
 end