karafka 1.4.12 → 2.2.10

data/lib/karafka/setup/config.rb

@@ -12,19 +12,54 @@ module Karafka
     #   enough and will still keep the code simple
     # @see Karafka::Setup::Configurators::Base for more details about configurators api
     class Config
-      extend Dry::Configurable
+      extend ::Karafka::Core::Configurable
 
-      # Contract for checking the config provided by the user
-      CONTRACT = Karafka::Contracts::Config.new.freeze
+      # Defaults for kafka settings, that will be overwritten only if not present already
+      KAFKA_DEFAULTS = {
+        # We emit the statistics by default, so all the instrumentation and web-ui work out of
+        # the box, without requiring users to take any extra actions aside from enabling.
+        'statistics.interval.ms': 5_000,
+        'client.software.name': 'karafka',
+        'client.software.version': [
+          "v#{Karafka::VERSION}",
+          "rdkafka-ruby-v#{Rdkafka::VERSION}",
+          "librdkafka-v#{Rdkafka::LIBRDKAFKA_VERSION}"
+        ].join('-')
+      }.freeze
 
-      private_constant :CONTRACT
+      # Contains settings that should not be used in production but make life easier in dev
+      KAFKA_DEV_DEFAULTS = {
+        # Will create non-existing topics automatically.
+        # Note that the broker needs to be configured with `auto.create.topics.enable=true`
+        # While it is not recommended in prod, it simplifies work in dev
+        'allow.auto.create.topics': 'true',
+        # We refresh the cluster state often as newly created topics in dev may not be detected
+        # fast enough. Fast enough means within reasonable time to provide decent user experience
+        # While it's only a one time thing for new topics, it can still be irritating to have to
+        # restart the process.
+        'topic.metadata.refresh.interval.ms': 5_000
+      }.freeze
+
+      private_constant :KAFKA_DEFAULTS, :KAFKA_DEV_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: ''
+      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)
@@ -33,195 +68,241 @@ 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 [String] should we start with the earliest possible offset or latest
+      # This will set the `auto.offset.reset` value unless present in the kafka scope
+      setting :initial_offset, default: 'earliest'
+      # options max_messages [Integer] how many messages do we want to fetch from Kafka in one go
+      setting :max_messages, default: 100
+      # option [Integer] number of milliseconds we can wait while fetching data
+      setting :max_wait_time, default: 1_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
+      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 (milliseconds)
+      setting :pause_timeout, default: 1_000
+      # option [Integer] what is the max timeout in case of an exponential backoff (milliseconds)
+      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
+      # option [Boolean] when set to true, Karafka will ensure that the routing topic naming
+      # convention is strict
+      # Disabling this may be needed in scenarios where we do not have control over topics names
+      # and/or we work with existing systems where we cannot change topics names.
+      setting :strict_topics_namespacing, default: true
+
+      # rdkafka default options
+      # @see https://github.com/edenhill/librdkafka/blob/master/CONFIGURATION.md
+      setting :kafka, default: {}
+
+      # Admin specific settings.
+      #
+      # Since admin operations are often specific, they may require specific librdkafka settings
+      # or other settings that are unique to admin.
+      setting :admin do
+        # Specific kafka settings that are tuned to operate within the Admin.
+        #
+        # Please do not change them unless you know what you are doing as their misconfiguration
+        # may cause Admin API to misbehave
+        # option [Hash] extra changes to the default root kafka settings
+        setting :kafka, default: {
+          # We want to know when there is no more data not to end up with an endless loop
+          'enable.partition.eof': true,
+          # Do not publish statistics from admin as they are not relevant
+          'statistics.interval.ms': 0,
+          # Fetch at most 5 MBs when using admin
+          'fetch.message.max.bytes': 5 * 1_048_576,
+          # Do not commit offset automatically, this prevents offset tracking for operations
+          # involving a consumer instance
+          'enable.auto.commit': false,
+          # Make sure that topic metadata lookups do not create topics accidentally
+          'allow.auto.create.topics': false
+        }
+
+        # option [String] default name for the admin consumer group. Please note, that this is a
+        # subject to be remapped by the consumer mapper as any other consumer group in the routes
+        setting :group_id, default: 'karafka_admin'
+
+        # option max_wait_time [Integer] We wait only for this amount of time before raising error
+        # as we intercept this error and retry after checking that the operation was finished or
+        # failed using external factor.
+        setting :max_wait_time, default: 1_000
+
+        # How many times should be try. 1 000 ms x 60 => 60 seconds wait in total and then we give
+        # up on pending operations
+        setting :max_attempts, default: 60
       end
 
-      # Namespace for internal settings that should not be modified
-      # It's a temporary step to "declassify" several things internally before we move to a
-      # non global state
+      # Namespace for internal settings that should not be modified directly
       setting :internal do
-        # option routing_builder [Karafka::Routing::Builder] builder instance
-        setting :routing_builder, default: Routing::Builder.new
         # option status [Karafka::Status] app status
         setting :status, default: Status.new
         # option process [Karafka::Process] process status
         # @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]
+
+        # Namespace for CLI related settings
+        setting :cli do
+          # option contract [Object] cli setup validation contract (in the context of options and
+          # topics)
+          setting :contract, default: Contracts::ServerCliOptions.new
+        end
+
+        setting :routing do
+          # option builder [Karafka::Routing::Builder] builder instance
+          setting :builder, default: Routing::Builder.new
+          # option subscription_groups_builder [Routing::SubscriptionGroupsBuilder] subscription
+          #   group builder
+          setting :subscription_groups_builder, default: Routing::SubscriptionGroupsBuilder.new
+
+          # Internally assigned list of limits on routings active for the current process
+          # This can be altered by the CLI command
+          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
+          # option jobs_builder [Object] jobs builder we want to use
+          setting :jobs_builder, default: Processing::JobsBuilder.new
+          # option coordinator [Class] work coordinator we want to user for processing coordination
+          setting :coordinator_class, default: Processing::Coordinator
+          # option partitioner_class [Class] partitioner we use against a batch of data
+          setting :partitioner_class, default: Processing::Partitioner
+          # option strategy_selector [Object] processing strategy selector to be used
+          setting :strategy_selector, default: Processing::StrategySelector.new
+          # option expansions_selector [Object] processing expansions selector to be used
+          setting :expansions_selector, default: Processing::ExpansionsSelector.new
+        end
+
+        # Things related to operating on messages
+        setting :messages do
+          # Parser is used to convert raw payload prior to deserialization
+          setting :parser, default: Messages::Parser.new
+        end
+
+        # Karafka components for ActiveJob
+        setting :active_job do
+          # option dispatcher [Karafka::ActiveJob::Dispatcher] default dispatcher for ActiveJob
+          setting :dispatcher, default: ActiveJob::Dispatcher.new
+          # option job_options_contract [Karafka::Contracts::JobOptionsContract] contract for
+          #   ensuring, that extra job options defined are valid
+          setting :job_options_contract, default: ActiveJob::JobOptionsContract.new
+          # option consumer [Class] consumer class that should be used to consume ActiveJob data
+          setting :consumer_class, default: ActiveJob::Consumer
+        end
       end
 
+      # This will load all the defaults that can be later overwritten.
+      # Thanks to that we have an initial state out of the box.
+      configure
+
       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) }
-        end
+        # @param block [Proc] block we want to execute with the config instance
+        def setup(&block)
+          # Will prepare and verify license if present
+          Licenser.prepare_and_verify(config.license)
+
+          # Pre-setup configure all routing features that would need this
+          Routing::Features::Base.pre_setup_all(config)
+
+          # Will configure all the pro components
+          # This needs to happen before end user configuration as the end user may overwrite some
+          # of the pro defaults with custom components
+          Pro::Loader.pre_setup_all(config) if Karafka.pro?
+
+          configure(&block)
+          merge_kafka_defaults!(config)
+
+          Contracts::Config.new.validate!(config.to_h)
+
+          configure_components
 
-        # 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) }
+          # Refreshes the references that are cached that might have been changed by the config
+          ::Karafka.refresh!
+
+          # Post-setup configure all routing features that would need this
+          Routing::Features::Base.post_setup_all(config)
+
+          # Runs things that need to be executed after config is defined and all the components
+          # are also configured
+          Pro::Loader.post_setup_all(config) if Karafka.pro?
+
+          Karafka::App.initialized!
         end
 
-        # Validate config based on the config contract
-        # @return [Boolean] true if configuration is valid
-        # @raise [Karafka::Errors::InvalidConfigurationError] raised when configuration
-        #   doesn't match with the config contract
-        def validate!
-          validation_result = CONTRACT.call(config.to_h)
+        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 [Karafka::Core::Configurable::Node] 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
 
-          return true if validation_result.success?
+          # Use Karafka client_id as kafka client id if not set
+          config.kafka[:'client.id'] ||= config.client_id
+
+          return if Karafka::App.env.production?
+
+          KAFKA_DEV_DEFAULTS.each do |key, value|
+            next if config.kafka.key?(key)
+
+            config.kafka[key] = value
+          end
+        end
 
-          raise Errors::InvalidConfigurationError, validation_result.errors.to_h
+        # 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 = AttributesMap.producer(config.kafka.dup)
+            producer_config.logger = config.logger
+          end
         end
       end
     end

data/lib/karafka/status.rb

@@ -3,15 +3,43 @@
 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!
+      # will no longer pickup any work, but current work will be finished
+      quieting: :quiet!,
+      # no work is happening but we keep process with the assignments running
+      quiet: :quieted!,
+      # shutdown started
+      stopping: :stop!,
+      # all things are done and most of the things except critical are closed
+      stopped: :stopped!,
+      # immediately after this process exists
+      terminated: :terminate!
     }.freeze
 
-    private_constant :STATES
+    # Mutex to ensure that state transitions are thread-safe
+    MUTEX = Mutex.new
+
+    private_constant :MUTEX
+
+    # By default we are in the initializing state
+    def initialize
+      initialize!
+    end
+
+    # @return [String] stringified current app status
+    def to_s
+      @status.to_s
+    end
+
+    # Resets the status state
+    # This is used mostly in the integration suite
+    def reset!
+      @status = :initializing
+    end
 
     STATES.each do |state, transition|
       define_method :"#{state}?" do
@@ -19,11 +47,30 @@ module Karafka
       end
 
       define_method transition do
-        @status = state
-        # 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
+        MUTEX.synchronize do
+          # Do not allow reverse state transitions (we always go one way) or transition to the same
+          # state as currently
+          return if @status && STATES.keys.index(state) <= STATES.keys.index(@status)
+
+          @status = state
+
+          # Skip on creation (initializing)
+          # We skip as during this state we do not have yet a monitor
+          return if initializing?
+
+          Karafka.monitor.instrument("app.#{state}")
+        end
       end
     end
+
+    # @return [Boolean] true if we are in any of the status that would indicate we should no longer
+    #   process incoming data. It is a meta status built from others and not a separate state in
+    #   the sense of a state machine
+    def done?
+      # Short-track for the most common case not to invoke all others on normal execution
+      return false if running?
+
+      stopping? || stopped? || quieting? || quiet? || terminated?
+    end
   end
 end

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 revoked
+  # end
+
+  # Define here any teardown things you want when Karafka server stops
+  # def shutdown
+  # end
+end