nulogy_message_bus_consumer 0.3.0 → 0.5.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 2faa5c725b717d217b37a0c9019ecd3f3ae9d40885a2e3b1ce898e626bbb2d3c
-  data.tar.gz: 865137198c57cfd60bdea31c482a66240a284c34d4f1149033160c28fb2a6cef
+  metadata.gz: 1dc5baf6066f0cf264ecf41623c1269580c8c000bd456505b3114918048b8598
+  data.tar.gz: 1a2e52f75e766213995846d7214d8f4e5248f011071f7d8a3fed7fb6c0fe2f03
 SHA512:
-  metadata.gz: 82c5cf164680d27183b17b341b362c5f1ec70a26fecf79d4f8d982ea3506e107ac8e216604c5182469c627892f3c423336053a2502c8e10c574c85af5dd82402
-  data.tar.gz: d2109e0aefef9eecebfb57425316cf36dd7cdb92a4356253efb73271f2fed5e4c1ed59e56a46b44a2dc869dd3acb5285b88ae11d6d72b15f50a7506ed85058a2
+  metadata.gz: 02effdb3a64405b47af2b6b63d1f4be505db7dcb8e632c568645af473e307030e7e0ac6615c48dc097fefaa32987f76b539ad28cfc65f5a7de98727b09d97052
+  data.tar.gz: 93dde5b9d7a0c12c973f6b2516d2867cc376298417c0e3ae8529f63e52974d8e489bd833494951db8cbb418eea7df124c2df02481a85eee8b7c9a575eefc8ceb

data/Rakefile

@@ -18,13 +18,14 @@ APP_RAKEFILE = File.expand_path("spec/dummy/Rakefile", __dir__)
 load "rails/tasks/engine.rake"
 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: [:spec, :rubocop]
+require "standard/rake"
+task default: %i[spec standard]
 
 require "rake/release"
 
 Rake::Release::Task.load_all do |spec|
-  spec.version_tag = "nulogy_message_bus_producer-v#{spec.version}"
+  spec.version_tag = "nulogy_message_bus_consumer-v#{spec.version}"
 end

data/config/credentials/message-bus-us-east-1.key

@@ -0,0 +1 @@
+dfa19863b2709390893da4c2fb85579a

data/config/credentials/message-bus-us-east-1.yml.enc

@@ -0,0 +1 @@
+/WXdqUYePaHqAq3P0iTEsrLiMfRKzp2qYYh7K+q6LUNgi4eHNv0+SoLdI1bUNb9UaGvDPGNT3fCwAdkurk5Iud16ok3b4wD6yZ7UkfqbXqZaKH/dciQ5s63p9Hiuq1rbfcqoZ3KR1SXYAwvy8vNqbdwbAzz2N66B1wE5fNibZZlrWzXJjLReiTcNyxNbCPz6vwEwFF52RntuYlIJo4Nkm8vEk3No+HWrOkM8xptr5qApbd+RowLCLZ4/kcAMDB/XPiGobf0AOFv1NUR/9ChEy20usa8Fqd6HtEn4A25HnAC0uaN0K8ZRXjxhMpnXtfMBItn7yxyJ1ubjRZK5a1xVBRU7L/CVV9ZuIsqAHL1++gH5FBrEe83ZIUhN7AzngMDlOPKGiCLiZLrm18I1AEQrD7tJLyXos15AeAzj--cER8cN0iMLwu8Le+--FL7dhMTgr6xL6SkMnYKmeg==

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
@@ -31,7 +36,7 @@ module NulogyMessageBusConsumer
   end
 
   def logger
-    @@logger ||= NullLogger.new
+    @logger ||= NullLogger.new
   end
 
   def invoke_pipeline(*steps)
@@ -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
 

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/handlers/log_unprocessed_messages.rb

@@ -7,9 +7,10 @@ module NulogyMessageBusConsumer
 
       def call(message:, **_)
         return if ProcessedMessage.exists?(id: message.id)
+
         @logger.warn(JSON.dump(
           event: "unprocessed_message",
-          kafka_message: message.to_h,
+          kafka_message: message.to_h
         ))
       end
     end

data/lib/nulogy_message_bus_consumer/kafka_utils.rb

@@ -13,13 +13,14 @@ module NulogyMessageBusConsumer
     def wait_for(attempts: 100, interval: 0.1)
       attempts.times do
         break if yield
+
         sleep interval
       end
     end
 
     def every_message_until_none_are_left(consumer)
       Enumerator.new do |yielder|
-        while message = consumer.poll(250)
+        while (message = consumer.poll(250))
           yielder.yield(message)
         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

@@ -1,9 +1,5 @@
 module NulogyMessageBusConsumer
   class Message
-    def initialize(attrs = {})
-      attrs.each { |key, value| instance_variable_set("@#{key}", value) }
-    end
-
     attr_reader :event_data
     attr_reader :event_data_unparsed
     attr_reader :id
@@ -14,10 +10,20 @@ module NulogyMessageBusConsumer
     attr_reader :company_uuid
     attr_reader :timestamp
     attr_reader :topic
+    attr_reader :created_at
+
+    def initialize(attrs = {})
+      attrs.each { |key, value| instance_variable_set("@#{key}", value) }
+    end
 
     def self.from_kafka(kafka_message)
       envelope_data = JSON.parse(kafka_message.payload, symbolize_names: true)
-      event_data = JSON.parse(envelope_data[:event_json], symbolize_names: true) rescue {}
+      event_data =
+        begin
+          JSON.parse(envelope_data[:event_json], symbolize_names: true)
+        rescue
+          {}
+        end
 
       new(
         event_data: event_data,
@@ -30,6 +36,7 @@ module NulogyMessageBusConsumer
         company_uuid: envelope_data[:company_uuid] || envelope_data[:tenant_id],
         timestamp: kafka_message.timestamp,
         topic: kafka_message.topic,
+        created_at: envelope_data[:created_at]
       )
     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

@@ -9,7 +9,7 @@ module NulogyMessageBusConsumer
     end
 
     def insert(step, after:)
-      index = @steps.find_index { |step| step.is_a?(after) }
+      index = @steps.find_index { |s| s.is_a?(after) }
       @steps.insert(index + 1, step)
     end
 
@@ -25,7 +25,7 @@ module NulogyMessageBusConsumer
       @steps.reverse.reduce(last_step) do |composed_steps, previous_step|
         lambda do |**args|
           invoke_next = compose_with_merged_args(args, composed_steps)
-            previous_step.call(**args, &invoke_next)
+          previous_step.call(**args, &invoke_next)
         end
       end
     end
@@ -33,7 +33,10 @@ module NulogyMessageBusConsumer
     def compose_with_merged_args(existing_args, func)
       lambda do |**yielded_args|
         args_to_be_overridden = existing_args.keys & yielded_args.keys
-        raise "Cannot override existing argument(s): #{args_to_be_overridden.join(', ')}" if args_to_be_overridden.any?
+        if args_to_be_overridden.any?
+          raise "Cannot override existing argument(s): #{args_to_be_overridden.join(", ")}"
+        end
+
         func.call(**existing_args.merge(yielded_args))
       end
     end

data/lib/nulogy_message_bus_consumer/steps/commit_on_success.rb

@@ -7,6 +7,7 @@ module NulogyMessageBusConsumer
         raise_if_invalid(result)
 
         if result == :success
+          kafka_consumer.store_offset(message)
           kafka_consumer.commit
         else
           reconnect_to_reprocess_same_message(kafka_consumer)

data/lib/nulogy_message_bus_consumer/steps/connect_to_message_bus.rb

@@ -1,33 +1,52 @@
 module NulogyMessageBusConsumer
   module Steps
     class ConnectToMessageBus
-      def initialize(config, logger)
+      def initialize(config, logger, kafka_consumer: nil)
         @config = config
         @logger = logger
+        @kafka_consumer = kafka_consumer
       end
 
       def call(**_)
         @logger.info("Connecting to the MessageBus")
-        consumer = Rdkafka::Config.new(consumer_config).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 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
         }
+
+        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

@@ -1,28 +1,39 @@
 module NulogyMessageBusConsumer
   module Steps
     class LogMessages
-      def initialize(logger)
+      def initialize(logger, clock: Clock.new)
         @logger = logger
+        @clock = clock
       end
 
       def call(message:, **_)
         @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.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
         }))
 
         result
       end
+
+      # Debezium appears to be giving us nanos since epoch
+      # https://github.com/debezium/debezium/blob/5a115e902cdc1dc399ec02758dd1039a33e99bc2/debezium-core/src/main/java/io/debezium/jdbc/JdbcValueConverters.java#L237
+      def diff_millis(oldest_nanos, newest_millis)
+        old_millis = oldest_nanos / 1000
+
+        newest_millis - old_millis
+      end
     end
   end
 end