sidekiq 6.5.12 → 7.3.9

data/lib/sidekiq/config.rb

@@ -0,0 +1,311 @@
+# frozen_string_literal: true
+
+require "forwardable"
+
+require "set"
+require "sidekiq/redis_connection"
+
+module Sidekiq
+  # Sidekiq::Config represents the global configuration for an instance of Sidekiq.
+  class Config
+    extend Forwardable
+
+    DEFAULTS = {
+      labels: Set.new,
+      require: ".",
+      environment: nil,
+      concurrency: 5,
+      timeout: 25,
+      poll_interval_average: nil,
+      average_scheduled_poll_interval: 5,
+      on_complex_arguments: :raise,
+      iteration: {
+        max_job_runtime: nil,
+        retry_backoff: 0
+      },
+      error_handlers: [],
+      death_handlers: [],
+      lifecycle_events: {
+        startup: [],
+        quiet: [],
+        shutdown: [],
+        # triggers when we fire the first heartbeat on startup OR repairing a network partition
+        heartbeat: [],
+        # triggers on EVERY heartbeat call, every 10 seconds
+        beat: []
+      },
+      dead_max_jobs: 10_000,
+      dead_timeout_in_seconds: 180 * 24 * 60 * 60, # 6 months
+      reloader: proc { |&block| block.call },
+      backtrace_cleaner: ->(backtrace) { backtrace }
+    }
+
+    ERROR_HANDLER = ->(ex, ctx, cfg = Sidekiq.default_configuration) {
+      l = cfg.logger
+      l.warn(Sidekiq.dump_json(ctx)) unless ctx.empty?
+      l.warn("#{ex.class.name}: #{ex.message}")
+      unless ex.backtrace.nil?
+        backtrace = cfg[:backtrace_cleaner].call(ex.backtrace)
+        l.warn(backtrace.join("\n"))
+      end
+    }
+
+    def initialize(options = {})
+      @options = DEFAULTS.merge(options)
+      @options[:error_handlers] << ERROR_HANDLER if @options[:error_handlers].empty?
+      @directory = {}
+      @redis_config = {}
+      @capsules = {}
+    end
+
+    def_delegators :@options, :[], :[]=, :fetch, :key?, :has_key?, :merge!, :dig
+    attr_reader :capsules
+
+    def inspect
+      "#<#{self.class.name} @options=#{
+        @options.except(:lifecycle_events, :reloader, :death_handlers, :error_handlers).inspect
+      }>"
+    end
+
+    def to_json(*)
+      Sidekiq.dump_json(@options)
+    end
+
+    # LEGACY: edits the default capsule
+    # config.concurrency = 5
+    def concurrency=(val)
+      default_capsule.concurrency = Integer(val)
+    end
+
+    def concurrency
+      default_capsule.concurrency
+    end
+
+    def total_concurrency
+      capsules.each_value.sum(&:concurrency)
+    end
+
+    # Edit the default capsule.
+    # config.queues = %w( high default low )                 # strict
+    # config.queues = %w( high,3 default,2 low,1 )           # weighted
+    # config.queues = %w( feature1,1 feature2,1 feature3,1 ) # random
+    #
+    # With weighted priority, queue will be checked first (weight / total) of the time.
+    # high will be checked first (3/6) or 50% of the time.
+    # I'd recommend setting weights between 1-10. Weights in the hundreds or thousands
+    # are ridiculous and unnecessarily expensive. You can get random queue ordering
+    # by explicitly setting all weights to 1.
+    def queues=(val)
+      default_capsule.queues = val
+    end
+
+    def queues
+      default_capsule.queues
+    end
+
+    def client_middleware
+      @client_chain ||= Sidekiq::Middleware::Chain.new(self)
+      yield @client_chain if block_given?
+      @client_chain
+    end
+
+    def server_middleware
+      @server_chain ||= Sidekiq::Middleware::Chain.new(self)
+      yield @server_chain if block_given?
+      @server_chain
+    end
+
+    def default_capsule(&block)
+      capsule("default", &block)
+    end
+
+    # register a new queue processing subsystem
+    def capsule(name)
+      nm = name.to_s
+      cap = @capsules.fetch(nm) do
+        cap = Sidekiq::Capsule.new(nm, self)
+        @capsules[nm] = cap
+      end
+      yield cap if block_given?
+      cap
+    end
+
+    # All capsules must use the same Redis configuration
+    def redis=(hash)
+      @redis_config = @redis_config.merge(hash)
+    end
+
+    def redis_pool
+      Thread.current[:sidekiq_redis_pool] || Thread.current[:sidekiq_capsule]&.redis_pool || local_redis_pool
+    end
+
+    private def local_redis_pool
+      # this is our internal client/housekeeping pool. each capsule has its
+      # own pool for executing threads.
+      @redis ||= new_redis_pool(10, "internal")
+    end
+
+    def new_redis_pool(size, name = "unset")
+      # connection pool is lazy, it will not create connections unless you actually need them
+      # so don't be skimpy!
+      RedisConnection.create({size: size, logger: logger, pool_name: name}.merge(@redis_config))
+    end
+
+    def redis_info
+      redis do |conn|
+        conn.call("INFO") { |i| i.lines(chomp: true).map { |l| l.split(":", 2) }.select { |l| l.size == 2 }.to_h }
+      rescue RedisClientAdapter::CommandError => ex
+        # 2850 return fake version when INFO command has (probably) been renamed
+        raise unless /unknown command/.match?(ex.message)
+        {
+          "redis_version" => "9.9.9",
+          "uptime_in_days" => "9999",
+          "connected_clients" => "9999",
+          "used_memory_human" => "9P",
+          "used_memory_peak_human" => "9P"
+        }.freeze
+      end
+    end
+
+    def redis
+      raise ArgumentError, "requires a block" unless block_given?
+      redis_pool.with do |conn|
+        retryable = true
+        begin
+          yield conn
+        rescue RedisClientAdapter::BaseError => ex
+          # 2550 Failover can cause the server to become a replica, need
+          # to disconnect and reopen the socket to get back to the primary.
+          # 4495 Use the same logic if we have a "Not enough replicas" error from the primary
+          # 4985 Use the same logic when a blocking command is force-unblocked
+          # The same retry logic is also used in client.rb
+          if retryable && ex.message =~ /READONLY|NOREPLICAS|UNBLOCKED/
+            conn.close
+            retryable = false
+            retry
+          end
+          raise
+        end
+      end
+    end
+
+    # register global singletons which can be accessed elsewhere
+    def register(name, instance)
+      # logger.debug("register[#{name}] = #{instance}")
+      # Sidekiq Enterprise lazy registers a few services so we
+      # can't lock down this hash completely.
+      hash = @directory.dup
+      hash[name] = instance
+      @directory = hash.freeze
+      instance
+    end
+
+    # find a singleton
+    def lookup(name, default_class = nil)
+      # JNDI is just a fancy name for a hash lookup
+      @directory.fetch(name) do |key|
+        return nil unless default_class
+        register(key, default_class.new(self))
+      end
+    end
+
+    def freeze!
+      @directory.freeze
+      @options.freeze
+      true
+    end
+
+    ##
+    # Death handlers are called when all retries for a job have been exhausted and
+    # the job dies.  It's the notification to your application
+    # that this job will not succeed without manual intervention.
+    #
+    # Sidekiq.configure_server do |config|
+    #   config.death_handlers << ->(job, ex) do
+    #   end
+    # end
+    def death_handlers
+      @options[:death_handlers]
+    end
+
+    # How frequently Redis should be checked by a random Sidekiq process for
+    # scheduled and retriable jobs. Each individual process will take turns by
+    # waiting some multiple of this value.
+    #
+    # See sidekiq/scheduled.rb for an in-depth explanation of this value
+    def average_scheduled_poll_interval=(interval)
+      @options[:average_scheduled_poll_interval] = interval
+    end
+
+    # Register a proc to handle any error which occurs within the Sidekiq process.
+    #
+    #   Sidekiq.configure_server do |config|
+    #     config.error_handlers << proc {|ex,ctx_hash| MyErrorService.notify(ex, ctx_hash) }
+    #   end
+    #
+    # The default error handler logs errors to @logger.
+    def error_handlers
+      @options[:error_handlers]
+    end
+
+    # Register a block to run at a point in the Sidekiq lifecycle.
+    # :startup, :quiet or :shutdown are valid events.
+    #
+    #   Sidekiq.configure_server do |config|
+    #     config.on(:shutdown) do
+    #       puts "Goodbye cruel world!"
+    #     end
+    #   end
+    def on(event, &block)
+      raise ArgumentError, "Symbols only please: #{event}" unless event.is_a?(Symbol)
+      raise ArgumentError, "Invalid event name: #{event}" unless @options[:lifecycle_events].key?(event)
+      @options[:lifecycle_events][event] << block
+    end
+
+    def logger
+      @logger ||= Sidekiq::Logger.new($stdout, level: :info).tap do |log|
+        log.level = Logger::INFO
+        log.formatter = if ENV["DYNO"]
+          Sidekiq::Logger::Formatters::WithoutTimestamp.new
+        else
+          Sidekiq::Logger::Formatters::Pretty.new
+        end
+      end
+    end
+
+    def logger=(logger)
+      if logger.nil?
+        self.logger.level = Logger::FATAL
+        return
+      end
+
+      @logger = logger
+    end
+
+    private def parameter_size(handler)
+      target = handler.is_a?(Proc) ? handler : handler.method(:call)
+      target.parameters.size
+    end
+
+    # INTERNAL USE ONLY
+    def handle_exception(ex, ctx = {})
+      if @options[:error_handlers].size == 0
+        p ["!!!!!", ex]
+      end
+      @options[:error_handlers].each do |handler|
+        if parameter_size(handler) == 2
+          # TODO Remove in 8.0
+          logger.info { "DEPRECATION: Sidekiq exception handlers now take three arguments, see #{handler}" }
+          handler.call(ex, {_config: self}.merge(ctx))
+        else
+          handler.call(ex, ctx, self)
+        end
+      rescue Exception => e
+        l = logger
+        l.error "!!! ERROR HANDLER THREW AN ERROR !!!"
+        l.error e
+        l.error e.backtrace.join("\n") unless e.backtrace.nil?
+      end
+    end
+  end
+end

data/lib/sidekiq/deploy.rb

@@ -0,0 +1,64 @@
+# frozen_string_literal: true
+
+require "sidekiq/redis_connection"
+require "time"
+
+# This file is designed to be required within the user's
+# deployment script; it should need a bare minimum of dependencies.
+# Usage:
+#
+#   require "sidekiq/deploy"
+#   Sidekiq::Deploy.mark!("Some change")
+#
+# If you do not pass a label, Sidekiq will try to use the latest
+# git commit info.
+#
+
+module Sidekiq
+  class Deploy
+    MARK_TTL = 90 * 24 * 60 * 60 # 90 days
+
+    LABEL_MAKER = -> {
+      `git log -1 --format="%h %s"`.strip
+    }
+
+    def self.mark!(label = nil)
+      Sidekiq::Deploy.new.mark!(label: label)
+    end
+
+    def initialize(pool = Sidekiq::RedisConnection.create)
+      @pool = pool
+    end
+
+    def mark!(at: Time.now, label: nil)
+      label ||= LABEL_MAKER.call
+      # we need to round the timestamp so that we gracefully
+      # handle an very common error in marking deploys:
+      # having every process mark its deploy, leading
+      # to N marks for each deploy. Instead we round the time
+      # to the minute so that multiple marks within that minute
+      # will all naturally rollup into one mark per minute.
+      whence = at.utc
+      floor = Time.utc(whence.year, whence.month, whence.mday, whence.hour, whence.min, 0)
+      datecode = floor.strftime("%Y%m%d")
+      key = "#{datecode}-marks"
+      stamp = floor.iso8601
+
+      @pool.with do |c|
+        # only allow one deploy mark for a given label for the next minute
+        lock = c.set("deploylock-#{label}", stamp, "nx", "ex", "60")
+        if lock
+          c.multi do |pipe|
+            pipe.hsetnx(key, stamp, label)
+            pipe.expire(key, MARK_TTL)
+          end
+        end
+      end
+    end
+
+    def fetch(date = Time.now.utc.to_date)
+      datecode = date.strftime("%Y%m%d")
+      @pool.with { |c| c.hgetall("#{datecode}-marks") }
+    end
+  end
+end

data/lib/sidekiq/embedded.rb

@@ -0,0 +1,63 @@
+# frozen_string_literal: true
+
+require "sidekiq/component"
+require "sidekiq/launcher"
+require "sidekiq/metrics/tracking"
+
+module Sidekiq
+  class Embedded
+    include Sidekiq::Component
+
+    def initialize(config)
+      @config = config
+    end
+
+    def run
+      housekeeping
+      fire_event(:startup, reverse: false, reraise: true)
+      @launcher = Sidekiq::Launcher.new(@config, embedded: true)
+      @launcher.run
+      sleep 0.2 # pause to give threads time to spin up
+
+      logger.info "Sidekiq running embedded, total process thread count: #{Thread.list.size}"
+      logger.debug { Thread.list.map(&:name) }
+    end
+
+    def quiet
+      @launcher&.quiet
+    end
+
+    def stop
+      @launcher&.stop
+    end
+
+    private
+
+    def housekeeping
+      logger.info "Running in #{RUBY_DESCRIPTION}"
+      logger.info Sidekiq::LICENSE
+      logger.info "Upgrade to Sidekiq Pro for more features and support: https://sidekiq.org" unless defined?(::Sidekiq::Pro)
+
+      # touch the connection pool so it is created before we
+      # fire startup and start multithreading.
+      info = config.redis_info
+      ver = Gem::Version.new(info["redis_version"])
+      raise "You are connecting to Redis #{ver}, Sidekiq requires Redis 6.2.0 or greater" if ver < Gem::Version.new("6.2.0")
+
+      maxmemory_policy = info["maxmemory_policy"]
+      if maxmemory_policy != "noeviction"
+        logger.warn <<~EOM
+
+
+          WARNING: Your Redis instance will evict Sidekiq data under heavy load.
+          The 'noeviction' maxmemory policy is recommended (current policy: '#{maxmemory_policy}').
+          See: https://github.com/sidekiq/sidekiq/wiki/Using-Redis#memory
+
+        EOM
+      end
+
+      logger.debug { "Client Middleware: #{@config.default_capsule.client_middleware.map(&:klass).join(", ")}" }
+      logger.debug { "Server Middleware: #{@config.default_capsule.server_middleware.map(&:klass).join(", ")}" }
+    end
+  end
+end

data/lib/sidekiq/fetch.rb

@@ -2,6 +2,7 @@
 
 require "sidekiq"
 require "sidekiq/component"
+require "sidekiq/capsule"
 
 module Sidekiq # :nodoc:
   class BasicFetch
@@ -26,31 +27,28 @@ module Sidekiq # :nodoc:
       end
     }
 
-    def initialize(config)
-      raise ArgumentError, "missing queue list" unless config[:queues]
-      @config = config
-      @strictly_ordered_queues = !!@config[:strict]
-      @queues = @config[:queues].map { |q| "queue:#{q}" }
-      if @strictly_ordered_queues
-        @queues.uniq!
-        @queues << {timeout: TIMEOUT}
-      end
+    def initialize(cap)
+      raise ArgumentError, "missing queue list" unless cap.queues
+      @config = cap
+      @strictly_ordered_queues = cap.mode == :strict
+      @queues = config.queues.map { |q| "queue:#{q}" }
+      @queues.uniq! if @strictly_ordered_queues
     end
 
     def retrieve_work
       qs = queues_cmd
       # 4825 Sidekiq Pro with all queues paused will return an
-      # empty set of queues with a trailing TIMEOUT value.
-      if qs.size <= 1
+      # empty set of queues
+      if qs.size <= 0
         sleep(TIMEOUT)
         return nil
       end
 
-      queue, job = redis { |conn| conn.brpop(*qs) }
+      queue, job = redis { |conn| conn.blocking_call(TIMEOUT, "brpop", *qs, TIMEOUT) }
       UnitOfWork.new(queue, job, config) if queue
     end
 
-    def bulk_requeue(inprogress, options)
+    def bulk_requeue(inprogress)
       return if inprogress.empty?
 
       logger.debug { "Re-queueing terminated jobs" }
@@ -83,7 +81,6 @@ module Sidekiq # :nodoc:
       else
         permute = @queues.shuffle
         permute.uniq!
-        permute << {timeout: TIMEOUT}
         permute
       end
     end

data/lib/sidekiq/iterable_job.rb

@@ -0,0 +1,55 @@
+# frozen_string_literal: true
+
+require "sidekiq/job/iterable"
+
+# Iterable jobs are ones which provide a sequence to process using
+# `build_enumerator(*args, cursor: cursor)` and then process each
+# element of that sequence in `each_iteration(item, *args)`.
+#
+# The job is kicked off as normal:
+#
+#     ProcessUserSet.perform_async(123)
+#
+# but instead of calling `perform`, Sidekiq will call:
+#
+#     enum = ProcessUserSet#build_enumerator(123, cursor:nil)
+#
+# Your Enumerator must yield `(object, updated_cursor)` and
+# Sidekiq will call your `each_iteration` method:
+#
+#     ProcessUserSet#each_iteration(object, 123)
+#
+# After every iteration, Sidekiq will check for shutdown. If we are
+# stopping, the cursor will be saved to Redis and the job re-queued
+# to pick up the rest of the work upon restart. Your job will get
+# the updated_cursor so it can pick up right where it stopped.
+#
+#     enum = ProcessUserSet#build_enumerator(123, cursor: updated_cursor)
+#
+# The cursor object must be serializable to JSON.
+#
+# Note there are several APIs to help you build enumerators for
+# ActiveRecord Relations, CSV files, etc. See sidekiq/job/iterable/*.rb.
+module Sidekiq
+  module IterableJob
+    def self.included(base)
+      base.include Sidekiq::Job
+      base.include Sidekiq::Job::Iterable
+    end
+
+    # def build_enumerator(*args, cursor:)
+    # def each_iteration(item, *args)
+
+    # Your job can also define several callbacks during points
+    # in each job's lifecycle.
+    #
+    # def on_start
+    # def on_resume
+    # def on_stop
+    # def on_complete
+    # def around_iteration
+    #
+    # To keep things simple and compatible, this is the same
+    # API as the `sidekiq-iteration` gem.
+  end
+end

data/lib/sidekiq/job/interrupt_handler.rb

@@ -0,0 +1,24 @@
+# frozen_string_literal: true
+
+module Sidekiq
+  module Job
+    class InterruptHandler
+      include Sidekiq::ServerMiddleware
+
+      def call(instance, hash, queue)
+        yield
+      rescue Interrupted
+        logger.debug "Interrupted, re-queueing..."
+        c = Sidekiq::Client.new
+        c.push(hash)
+        raise Sidekiq::JobRetry::Skip
+      end
+    end
+  end
+end
+
+Sidekiq.configure_server do |config|
+  config.server_middleware do |chain|
+    chain.add Sidekiq::Job::InterruptHandler
+  end
+end

data/lib/sidekiq/job/iterable/active_record_enumerator.rb

@@ -0,0 +1,53 @@
+# frozen_string_literal: true
+
+module Sidekiq
+  module Job
+    module Iterable
+      # @api private
+      class ActiveRecordEnumerator
+        def initialize(relation, cursor: nil, **options)
+          @relation = relation
+          @cursor = cursor
+          @options = options
+        end
+
+        def records
+          Enumerator.new(-> { @relation.count }) do |yielder|
+            @relation.find_each(**@options, start: @cursor) do |record|
+              yielder.yield(record, record.id)
+            end
+          end
+        end
+
+        def batches
+          Enumerator.new(-> { @relation.count }) do |yielder|
+            @relation.find_in_batches(**@options, start: @cursor) do |batch|
+              yielder.yield(batch, batch.first.id)
+            end
+          end
+        end
+
+        def relations
+          Enumerator.new(-> { relations_size }) do |yielder|
+            # Convenience to use :batch_size for all the
+            # ActiveRecord batching methods.
+            options = @options.dup
+            options[:of] ||= options.delete(:batch_size)
+
+            @relation.in_batches(**options, start: @cursor) do |relation|
+              first_record = relation.first
+              yielder.yield(relation, first_record.id)
+            end
+          end
+        end
+
+        private
+
+        def relations_size
+          batch_size = @options[:batch_size] || 1000
+          (@relation.count + batch_size - 1) / batch_size # ceiling division
+        end
+      end
+    end
+  end
+end

data/lib/sidekiq/job/iterable/csv_enumerator.rb

@@ -0,0 +1,47 @@
+# frozen_string_literal: true
+
+module Sidekiq
+  module Job
+    module Iterable
+      # @api private
+      class CsvEnumerator
+        def initialize(csv)
+          unless defined?(CSV) && csv.instance_of?(CSV)
+            raise ArgumentError, "CsvEnumerator.new takes CSV object"
+          end
+
+          @csv = csv
+        end
+
+        def rows(cursor:)
+          @csv.lazy
+            .each_with_index
+            .drop(cursor || 0)
+            .to_enum { count_of_rows_in_file }
+        end
+
+        def batches(cursor:, batch_size: 100)
+          @csv.lazy
+            .each_slice(batch_size)
+            .with_index
+            .drop(cursor || 0)
+            .to_enum { (count_of_rows_in_file.to_f / batch_size).ceil }
+        end
+
+        private
+
+        def count_of_rows_in_file
+          filepath = @csv.path
+          return unless filepath
+
+          count = IO.popen(["wc", "-l", filepath]) do |out|
+            out.read.strip.to_i
+          end
+
+          count -= 1 if @csv.headers
+          count
+        end
+      end
+    end
+  end
+end