nulogy_message_bus_consumer 0.3.3 → 0.4.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: c47fb7fab95cd7d64e9654c0ce43fec64f77f8718331f8028afc6c18ab993f81
-  data.tar.gz: 3c656a7dea18d149f04ccfbc3dc4db04e9aa7b3cd6183be7b0a1a45337c3a975
+  metadata.gz: c7a2f6f3118c35e9f6b0f4cfebd87b96d0f17fbd391130953b560d61d642f73e
+  data.tar.gz: 6ff5ee3e40fd44c96592fdba4c78b47b3c87aa4c16f753ef8e9acbf5af1b429e
 SHA512:
-  metadata.gz: 04320f62a752bcfcc4bad93bd0611ba0f8945018dfb1680016ab739b70ba66cb7ebcf5062fb12f316745e59386714613c44f17f7feea4006a56b8625da119e23
-  data.tar.gz: c480ced5fb50feeb8d655cb8c0b45c19a12bd733088db6fe4bf78ce33d08a7051594eb86a011227d2cc1363ea058e1544435c318824a6c9549be2e32fb6f65ef
+  metadata.gz: 80ee9d751672ab78511a130e22926f0d9f2e6abeb5ed5447baa5866180a75387bad43b7e4b292d300a2853176d6ed24df701b268c43c6c4495d3397fb5b62415
+  data.tar.gz: 5aae3dc5540949b92e086c39b038dab2d2c2959535c777557279e70023e0ce76a6585881dfb293783aa55ebd9c1910b897f1f609ff257526bd7ebb4a8f393a18

data/Rakefile

@@ -21,9 +21,8 @@ load "rails/tasks/statistics.rake"
 require "rspec/core"
 require "rspec/core/rake_task"
 RSpec::Core::RakeTask.new(:spec)
-require "rubocop/rake_task"
-RuboCop::RakeTask.new
-task default: %i[spec rubocop]
+require "standard/rake"
+task default: %i[spec standard]
 
 require "rake/release"
 

data/lib/nulogy_message_bus_consumer.rb

@@ -1,10 +1,14 @@
+require "active_record/railtie"
+require "active_support/core_ext/time/zones"
 require "rdkafka"
 
 require "nulogy_message_bus_consumer/engine"
-
+require "nulogy_message_bus_consumer/clock"
 require "nulogy_message_bus_consumer/config"
+require "nulogy_message_bus_consumer/deployment/ecs"
 require "nulogy_message_bus_consumer/handlers/log_unprocessed_messages"
 require "nulogy_message_bus_consumer/kafka_utils"
+require "nulogy_message_bus_consumer/lag_tracker"
 require "nulogy_message_bus_consumer/message"
 require "nulogy_message_bus_consumer/null_logger"
 require "nulogy_message_bus_consumer/pipeline"
@@ -12,11 +16,12 @@ require "nulogy_message_bus_consumer/processed_message"
 require "nulogy_message_bus_consumer/steps/commit_on_success"
 require "nulogy_message_bus_consumer/steps/connect_to_message_bus"
 require "nulogy_message_bus_consumer/steps/deduplicate_messages"
+require "nulogy_message_bus_consumer/steps/log_consumer_lag"
 require "nulogy_message_bus_consumer/steps/log_messages"
-require "nulogy_message_bus_consumer/steps/monitor_replication_lag"
 require "nulogy_message_bus_consumer/steps/seek_beginning_of_topic"
 require "nulogy_message_bus_consumer/steps/stream_messages"
 require "nulogy_message_bus_consumer/steps/stream_messages_until_none_are_left"
+require "nulogy_message_bus_consumer/steps/supervise_consumer_lag"
 
 module NulogyMessageBusConsumer
   module_function
@@ -40,14 +45,21 @@ module NulogyMessageBusConsumer
 
   def recommended_consumer_pipeline(config: self.config, logger: self.logger)
     Pipeline.new([
-      # The first three are really system processing steps
+      # System processing/health steps.
+      # Note: that since they are before `StreamMessages`, they will only
+      # be called once, without any messages.
       Steps::ConnectToMessageBus.new(config, logger),
-      Steps::MonitorReplicationLag.new(logger),
+      Steps::LogConsumerLag.new(logger),
+      Steps::SuperviseConsumerLag.new(
+        logger,
+        check_interval_seconds: config.lag_check_interval_seconds,
+        tracker: LagTracker.new(failing_checks: config.lag_checks)
+      ),
       Steps::StreamMessages.new(logger),
       # Message processing steps start here.
       Steps::LogMessages.new(logger),
       Steps::CommitOnSuccess.new,
-      Steps::DeduplicateMessages.new(logger),
+      Steps::DeduplicateMessages.new(logger)
     ])
   end
 
@@ -56,7 +68,7 @@ module NulogyMessageBusConsumer
       Steps::ConnectToMessageBus.new(config, logger),
       Steps::SeekBeginningOfTopic.new,
       Steps::StreamMessagesUntilNoneAreLeft.new(logger),
-      Handlers::LogUnprocessedMessages.new(logger),
+      Handlers::LogUnprocessedMessages.new(logger)
     ])
   end
 end

data/lib/nulogy_message_bus_consumer/clock.rb

@@ -0,0 +1,13 @@
+module NulogyMessageBusConsumer
+  # Note: Since this calls Time.zone, it is NOt thread-safe
+  class Clock
+    def now
+      Time.zone.now.to_datetime
+    end
+
+    # milliseconds since epoch
+    def ms
+      now.strftime("%Q").to_i
+    end
+  end
+end

data/lib/nulogy_message_bus_consumer/config.rb

@@ -1,11 +1,19 @@
 module NulogyMessageBusConsumer
   class Config
-    attr_accessor :consumer_group_id
-    attr_accessor :bootstrap_servers
-    attr_accessor :topic_name
+    attr_accessor :bootstrap_servers,
+      :client_id,
+      :consumer_group_id,
+      :lag_check_interval_seconds,
+      :lag_checks,
+      :topic_name
 
     def initialize(options = {})
-      update(options)
+      defaults = {
+        lag_check_interval_seconds: 20,
+        lag_checks: 6
+      }
+
+      update(defaults.merge(options))
     end
 
     def update(options = {})

data/lib/nulogy_message_bus_consumer/deployment/ecs.rb

@@ -0,0 +1,23 @@
+module NulogyMessageBusConsumer
+  module Deployment
+    module ECS
+      module_function
+
+      # Try to get the TaskID from metadata server:
+      # https://docs.aws.amazon.com/AmazonECS/latest/developerguide/task-metadata-endpoint-v4.html
+      # Otherwise, return nil
+      def task_id
+        data = `curl --silent "$ECS_CONTAINER_METADATA_URI_V4/task"`
+
+        return if data.empty?
+
+        json = JSON.parse(data)
+        arn = json["TaskARN"]
+
+        return unless arn
+
+        arn.split("/").last
+      end
+    end
+  end
+end

data/lib/nulogy_message_bus_consumer/lag_tracker.rb

@@ -0,0 +1,53 @@
+require "set"
+
+module NulogyMessageBusConsumer
+  # Keeps track of how many times a topic's partition has not changed (non-zero) lag between update calls.
+  class LagTracker
+    attr_reader :failing_checks
+
+    def initialize(failing_checks: 3)
+      @failing_checks = failing_checks
+      @tracked = Hash.new { |h, topic| h[topic] = {} }
+      @failed = Hash.new { |h, topic| h[topic] = Set.new }
+    end
+
+    def update(topic_partitions)
+      topic_partitions.each_pair do |topic, partitions|
+        partitions.each_pair do |partition, value|
+          update_topic_partition(topic, partition, value)
+        end
+      end
+    end
+
+    def failing?
+      @failed.any?
+    end
+
+    def failed
+      @failed.transform_values { |v| v.to_a.sort }
+    end
+
+    private
+
+    def update_topic_partition(topic, partition, value)
+      current_value, count = @tracked.dig(topic, partition)
+
+      new_value, new_count =
+        if current_value == value && !value.zero?
+          [current_value, count + 1]
+        else
+          [value, 0]
+        end
+
+      @tracked[topic][partition] = [new_value, new_count]
+
+      if new_count >= @failing_checks
+        @failed[topic] << partition
+      end
+    end
+
+    def exists?(topic, partition)
+      @tracked.dig(topic, partition)
+    end
+  end
+end

data/lib/nulogy_message_bus_consumer/message.rb

@@ -21,7 +21,7 @@ module NulogyMessageBusConsumer
       event_data =
         begin
           JSON.parse(envelope_data[:event_json], symbolize_names: true)
-        rescue StandardError
+        rescue
           {}
         end
 

data/lib/nulogy_message_bus_consumer/null_logger.rb

@@ -1,9 +1,12 @@
 module NulogyMessageBusConsumer
   class NullLogger
-    def info(*_) end
+    def info(*_)
+    end
 
-    def error(*_) end
+    def error(*_)
+    end
 
-    def warn(*_) end
+    def warn(*_)
+    end
   end
 end

data/lib/nulogy_message_bus_consumer/pipeline.rb

@@ -34,7 +34,7 @@ module NulogyMessageBusConsumer
       lambda do |**yielded_args|
         args_to_be_overridden = existing_args.keys & yielded_args.keys
         if args_to_be_overridden.any?
-          raise "Cannot override existing argument(s): #{args_to_be_overridden.join(', ')}"
+          raise "Cannot override existing argument(s): #{args_to_be_overridden.join(", ")}"
         end
 
         func.call(**existing_args.merge(yielded_args))

data/lib/nulogy_message_bus_consumer/steps/connect_to_message_bus.rb

@@ -9,31 +9,44 @@ module NulogyMessageBusConsumer
 
       def call(**_)
         @logger.info("Connecting to the MessageBus")
-        consumer = build_consumer
         @logger.info("Using consumer group id: #{@config.consumer_group_id}")
 
-        consumer.subscribe(@config.topic_name)
-        @logger.info("Listening for kafka messages on topic #{@config.topic_name}")
+        subscribe
+
+        trap("TERM") { kafka_consumer.close }
 
-        trap("TERM") { consumer.close }
+        wait_for_assignment
 
-        KafkaUtils.wait_for_assignment(consumer)
-        yield(kafka_consumer: consumer)
+        yield(kafka_consumer: kafka_consumer)
       end
 
       private
 
-      def build_consumer
-        @kafka_consumer || Rdkafka::Config.new(consumer_config).consumer
+      def kafka_consumer
+        @kafka_consumer ||= Rdkafka::Config.new(consumer_config).consumer
       end
 
       def consumer_config
-        {
+        config = {
           "bootstrap.servers": @config.bootstrap_servers,
           "enable.auto.commit": false,
           "group.id": @config.consumer_group_id,
-          "enable.auto.offset.store": false,
+          "enable.auto.offset.store": false
         }
+
+        config["client.id"] = @config.client_id if @config.client_id
+
+        config
+      end
+
+      def subscribe
+        kafka_consumer.subscribe(@config.topic_name)
+        @logger.info("Listening for kafka messages on topic #{@config.topic_name}")
+      end
+
+      def wait_for_assignment
+        KafkaUtils.wait_for_assignment(kafka_consumer)
+        @logger.info("Connected as client: #{kafka_consumer.member_id}")
       end
     end
   end

data/lib/nulogy_message_bus_consumer/steps/deduplicate_messages.rb

@@ -41,7 +41,7 @@ module NulogyMessageBusConsumer
       def log_duplicate(message)
         @logger.warn(JSON.dump({
           event: "duplicate_message_detected",
-          kafka_message_id: message.id,
+          kafka_message_id: message.id
         }))
       end
     end

data/lib/nulogy_message_bus_consumer/steps/{monitor_replication_lag.rb → log_consumer_lag.rb}

@@ -1,6 +1,6 @@
 module NulogyMessageBusConsumer
   module Steps
-    class MonitorReplicationLag
+    class LogConsumerLag
       def initialize(logger)
         @logger = logger
       end
@@ -22,9 +22,9 @@ module NulogyMessageBusConsumer
 
             @logger.info(JSON.dump({
               event: "consumer_lag",
-              topics: Calculator.add_max_lag(lag_per_topic),
+              topics: Calculator.add_max_lag(lag_per_topic)
             }))
-            STDOUT.flush
+            $stdout.flush
 
             sleep 60
           end

data/lib/nulogy_message_bus_consumer/steps/log_messages.rb

@@ -1,12 +1,5 @@
 module NulogyMessageBusConsumer
   module Steps
-    class Clock
-      # milliseconds since epoch
-      def now
-        Time.zone.now.to_datetime.strftime("%Q").to_i
-      end
-    end
-
     class LogMessages
       def initialize(logger, clock: Clock.new)
         @logger = logger
@@ -17,18 +10,18 @@ module NulogyMessageBusConsumer
         @logger.info(JSON.dump({
           event: "message_received",
           kafka_message_id: message.id,
-          message: "Received #{message.id}",
+          message: "Received #{message.id}"
         }))
 
         result = yield
 
-        millis = diff_millis(message.created_at, @clock.now)
+        millis = diff_millis(message.created_at, @clock.ms)
         @logger.info(JSON.dump({
           event: "message_processed",
           kafka_message_id: message.id,
-          message: "Processed #{message.id}",
+          message: "Processed #{message.id} (#{message.topic}##{message.partition}@#{message.offset})",
           result: result,
-          time_to_processed: millis,
+          time_to_processed: millis
         }))
 
         result

data/lib/nulogy_message_bus_consumer/steps/stream_messages.rb

@@ -12,11 +12,11 @@ module NulogyMessageBusConsumer
             kafka_message: kafka_message
           )
         end
-      rescue StandardError => e
+      rescue => e
         @logger.error(JSON.dump({
           event: "message_processing_errored",
           class: e.class,
-          message: e.message,
+          message: e.message
         }))
 
         raise

data/lib/nulogy_message_bus_consumer/steps/stream_messages_until_none_are_left.rb

@@ -12,11 +12,11 @@ module NulogyMessageBusConsumer
             kafka_message: kafka_message
           )
         end
-      rescue StandardError => e
+      rescue => e
         @logger.error(JSON.dump({
           event: "message_processing_errored",
           class: e.class,
-          message: e.message,
+          message: e.message
         }))
 
         raise

data/lib/nulogy_message_bus_consumer/steps/supervise_consumer_lag.rb

@@ -0,0 +1,76 @@
+module NulogyMessageBusConsumer
+  module Steps
+    # Supervises the consumer's lag.
+    #
+    # If a partition's lag is non-zero and does not change for an extended period
+    # of time, then kill the main thread.
+    #
+    # That period of time is check_interval_seconds * LagTracker#failing_checks
+    # With the defaults, that would be 20 * 6 ~ 120 seconds = 2 minutes.
+    #
+    # Note that this strategy may not work for a busy integration.
+    # Consumer lag monitoring should alert in that case.
+    # However, this strategy may help alleviate alerts for low traffic or off-peak
+    # environments.
+    #
+    # We've come across cases where the consumer lag is still being logged,
+    # messages are being processed, but the consumer is not consuming messages
+    # in particular topics.
+    #
+    # Killing the main thread causes ECS to restart the task.
+    class SuperviseConsumerLag
+      def initialize(logger, tracker: NulogyMessageBusConsumer::LagTracker.new(failing_checks: 6), killable: nil, check_interval_seconds: 20)
+        @logger = logger
+        @tracker = tracker
+        @killable = killable
+        @check_interval_seconds = check_interval_seconds
+      end
+
+      def call(kafka_consumer:, **_)
+        @consumer = kafka_consumer
+        @killable ||= Thread.current
+
+        run
+
+        yield
+      end
+
+      private
+
+      def run
+        Thread.abort_on_exception = true
+
+        Thread.new do
+          NulogyMessageBusConsumer::KafkaUtils.wait_for_assignment(@consumer)
+
+          loop do
+            @tracker.update(@consumer.lag(@consumer.committed))
+
+            if @tracker.failing?
+              log_failed_partitions
+
+              @killable.kill
+              Thread.current.exit
+            end
+
+            sleep @check_interval_seconds
+          end
+        end
+      end
+
+      def log_failed_partitions
+        seconds = @check_interval_seconds * @tracker.failing_checks
+        failed = @tracker
+          .failed
+          .map { |topic, partitions| "#{topic}: #{partitions.join(",")}" }
+          .join(", ")
+
+        @logger.warn(JSON.dump({
+          event: "message_processing_warning",
+          message: "Assigned partition lag has not changed in #{seconds} seconds: #{failed}"
+        }))
+        $stdout.flush
+      end
+    end
+  end
+end

data/lib/nulogy_message_bus_consumer/version.rb

@@ -1,3 +1,3 @@
 module NulogyMessageBusConsumer
-  VERSION = "0.3.3"
+  VERSION = "0.4.0"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: nulogy_message_bus_consumer
 version: !ruby/object:Gem::Version
-  version: 0.3.3
+  version: 0.4.0
 platform: ruby
 authors:
 - Nulogy
-autorequire: 
+autorequire:
 bindir: bin
 cert_chain: []
-date: 2020-12-02 00:00:00.000000000 Z
+date: 2021-03-30 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: activerecord
@@ -142,14 +142,14 @@ dependencies:
     requirements:
     - - '='
       - !ruby/object:Gem::Version
-        version: 1.2.1
+        version: 1.3.0
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - '='
       - !ruby/object:Gem::Version
-        version: 1.2.1
+        version: 1.3.0
 - !ruby/object:Gem::Dependency
   name: rspec
   requirement: !ruby/object:Gem::Requirement
@@ -193,19 +193,19 @@ dependencies:
       - !ruby/object:Gem::Version
         version: 4.0.1
 - !ruby/object:Gem::Dependency
-  name: rubocop
+  name: standard
   requirement: !ruby/object:Gem::Requirement
     requirements:
     - - '='
       - !ruby/object:Gem::Version
-        version: 0.81.0
+        version: 0.11.0
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - '='
       - !ruby/object:Gem::Version
-        version: 0.81.0
+        version: 0.11.0
 - !ruby/object:Gem::Dependency
   name: rubocop-rails
   requirement: !ruby/object:Gem::Requirement
@@ -234,7 +234,7 @@ dependencies:
     - - '='
       - !ruby/object:Gem::Version
         version: 1.38.1
-description: 
+description:
 email:
 - tass@nulogy.com
 executables: []
@@ -247,10 +247,13 @@ files:
 - config/routes.rb
 - db/migrate/20200509095105_create_message_bus_processed_messages.rb
 - lib/nulogy_message_bus_consumer.rb
+- lib/nulogy_message_bus_consumer/clock.rb
 - lib/nulogy_message_bus_consumer/config.rb
+- lib/nulogy_message_bus_consumer/deployment/ecs.rb
 - lib/nulogy_message_bus_consumer/engine.rb
 - lib/nulogy_message_bus_consumer/handlers/log_unprocessed_messages.rb
 - lib/nulogy_message_bus_consumer/kafka_utils.rb
+- lib/nulogy_message_bus_consumer/lag_tracker.rb
 - lib/nulogy_message_bus_consumer/message.rb
 - lib/nulogy_message_bus_consumer/null_logger.rb
 - lib/nulogy_message_bus_consumer/pipeline.rb
@@ -258,18 +261,19 @@ files:
 - lib/nulogy_message_bus_consumer/steps/commit_on_success.rb
 - lib/nulogy_message_bus_consumer/steps/connect_to_message_bus.rb
 - lib/nulogy_message_bus_consumer/steps/deduplicate_messages.rb
+- lib/nulogy_message_bus_consumer/steps/log_consumer_lag.rb
 - lib/nulogy_message_bus_consumer/steps/log_messages.rb
-- lib/nulogy_message_bus_consumer/steps/monitor_replication_lag.rb
 - lib/nulogy_message_bus_consumer/steps/seek_beginning_of_topic.rb
 - lib/nulogy_message_bus_consumer/steps/stream_messages.rb
 - lib/nulogy_message_bus_consumer/steps/stream_messages_until_none_are_left.rb
+- lib/nulogy_message_bus_consumer/steps/supervise_consumer_lag.rb
 - lib/nulogy_message_bus_consumer/version.rb
 - lib/tasks/engine/message_bus_consumer.rake
 homepage: https://github.com/nulogy/message-bus/tree/master/gems/nulogy_message_bus_consumer
 licenses: []
 metadata:
   allowed_push_host: https://rubygems.org/
-post_install_message: 
+post_install_message:
 rdoc_options: []
 require_paths:
 - lib
@@ -285,7 +289,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
       version: '0'
 requirements: []
 rubygems_version: 3.0.3
-signing_key: 
+signing_key:
 specification_version: 4
 summary: Code for accessing the Nulogy Message Bus
 test_files: []