sidekiq 4.2.10 → 6.5.7

data/lib/sidekiq/api.rb

@@ -1,11 +1,36 @@
-# encoding: utf-8
 # frozen_string_literal: true
-require 'sidekiq'
+
+require "sidekiq"
+
+require "zlib"
+require "set"
+require "base64"
+
+if ENV["SIDEKIQ_METRICS_BETA"]
+  require "sidekiq/metrics/deploy"
+  require "sidekiq/metrics/query"
+end
+
+#
+# Sidekiq's Data API provides a Ruby object model on top
+# of Sidekiq's runtime data in Redis. This API should never
+# be used within application code for business logic.
+#
+# The Sidekiq server process never uses this API: all data
+# manipulation is done directly for performance reasons to
+# ensure we are using Redis as efficiently as possible at
+# every callsite.
+#
 
 module Sidekiq
+  # Retrieve runtime statistics from Redis regarding
+  # this Sidekiq cluster.
+  #
+  #   stat = Sidekiq::Stats.new
+  #   stat.processed
   class Stats
     def initialize
-      fetch_stats!
+      fetch_stats_fast!
     end
 
     def processed
@@ -48,56 +73,82 @@ module Sidekiq
       Sidekiq::Stats::Queues.new.lengths
     end
 
-    def fetch_stats!
-      pipe1_res = Sidekiq.redis do |conn|
-        conn.pipelined do
-          conn.get('stat:processed'.freeze)
-          conn.get('stat:failed'.freeze)
-          conn.zcard('schedule'.freeze)
-          conn.zcard('retry'.freeze)
-          conn.zcard('dead'.freeze)
-          conn.scard('processes'.freeze)
-          conn.lrange('queue:default'.freeze, -1, -1)
-          conn.smembers('processes'.freeze)
-          conn.smembers('queues'.freeze)
+    # O(1) redis calls
+    # @api private
+    def fetch_stats_fast!
+      pipe1_res = Sidekiq.redis { |conn|
+        conn.pipelined do |pipeline|
+          pipeline.get("stat:processed")
+          pipeline.get("stat:failed")
+          pipeline.zcard("schedule")
+          pipeline.zcard("retry")
+          pipeline.zcard("dead")
+          pipeline.scard("processes")
+          pipeline.lrange("queue:default", -1, -1)
         end
-      end
+      }
 
-      pipe2_res = Sidekiq.redis do |conn|
-        conn.pipelined do
-          pipe1_res[7].each {|key| conn.hget(key, 'busy'.freeze) }
-          pipe1_res[8].each {|queue| conn.llen("queue:#{queue}") }
+      default_queue_latency = if (entry = pipe1_res[6].first)
+        job = begin
+          Sidekiq.load_json(entry)
+        rescue
+          {}
         end
+        now = Time.now.to_f
+        thence = job["enqueued_at"] || now
+        now - thence
+      else
+        0
       end
 
-      s = pipe1_res[7].size
-      workers_size = pipe2_res[0...s].map(&:to_i).inject(0, &:+)
-      enqueued     = pipe2_res[s..-1].map(&:to_i).inject(0, &:+)
-
-      default_queue_latency = if (entry = pipe1_res[6].first)
-                                job = Sidekiq.load_json(entry)
-                                now = Time.now.to_f
-                                thence = job['enqueued_at'.freeze] || now
-                                now - thence
-                              else
-                                0
-                              end
       @stats = {
-        processed:             pipe1_res[0].to_i,
-        failed:                pipe1_res[1].to_i,
-        scheduled_size:        pipe1_res[2],
-        retry_size:            pipe1_res[3],
-        dead_size:             pipe1_res[4],
-        processes_size:        pipe1_res[5],
-
-        default_queue_latency: default_queue_latency,
-        workers_size:          workers_size,
-        enqueued:              enqueued
+        processed: pipe1_res[0].to_i,
+        failed: pipe1_res[1].to_i,
+        scheduled_size: pipe1_res[2],
+        retry_size: pipe1_res[3],
+        dead_size: pipe1_res[4],
+        processes_size: pipe1_res[5],
+
+        default_queue_latency: default_queue_latency
+      }
+    end
+
+    # O(number of processes + number of queues) redis calls
+    # @api private
+    def fetch_stats_slow!
+      processes = Sidekiq.redis { |conn|
+        conn.sscan_each("processes").to_a
       }
+
+      queues = Sidekiq.redis { |conn|
+        conn.sscan_each("queues").to_a
+      }
+
+      pipe2_res = Sidekiq.redis { |conn|
+        conn.pipelined do |pipeline|
+          processes.each { |key| pipeline.hget(key, "busy") }
+          queues.each { |queue| pipeline.llen("queue:#{queue}") }
+        end
+      }
+
+      s = processes.size
+      workers_size = pipe2_res[0...s].sum(&:to_i)
+      enqueued = pipe2_res[s..-1].sum(&:to_i)
+
+      @stats[:workers_size] = workers_size
+      @stats[:enqueued] = enqueued
+      @stats
     end
 
+    # @api private
+    def fetch_stats!
+      fetch_stats_fast!
+      fetch_stats_slow!
+    end
+
+    # @api private
     def reset(*stats)
-      all   = %w(failed processed)
+      all = %w[failed processed]
       stats = stats.empty? ? all : all & stats.flatten.compact.map(&:to_s)
 
       mset_args = []
@@ -113,66 +164,62 @@ module Sidekiq
     private
 
     def stat(s)
-      @stats[s]
+      fetch_stats_slow! if @stats[s].nil?
+      @stats[s] || raise(ArgumentError, "Unknown stat #{s}")
     end
 
     class Queues
       def lengths
         Sidekiq.redis do |conn|
-          queues = conn.smembers('queues'.freeze)
+          queues = conn.sscan_each("queues").to_a
 
-          lengths = conn.pipelined do
+          lengths = conn.pipelined { |pipeline|
             queues.each do |queue|
-              conn.llen("queue:#{queue}")
+              pipeline.llen("queue:#{queue}")
             end
-          end
+          }
 
-          i = 0
-          array_of_arrays = queues.inject({}) do |memo, queue|
-            memo[queue] = lengths[i]
-            i += 1
-            memo
-          end.sort_by { |_, size| size }
-
-          Hash[array_of_arrays.reverse]
+          array_of_arrays = queues.zip(lengths).sort_by { |_, size| -size }
+          array_of_arrays.to_h
         end
       end
     end
 
     class History
       def initialize(days_previous, start_date = nil)
+        # we only store five years of data in Redis
+        raise ArgumentError if days_previous < 1 || days_previous > (5 * 365)
         @days_previous = days_previous
         @start_date = start_date || Time.now.utc.to_date
       end
 
       def processed
-        date_stat_hash("processed")
+        @processed ||= date_stat_hash("processed")
       end
 
       def failed
-        date_stat_hash("failed")
+        @failed ||= date_stat_hash("failed")
       end
 
       private
 
       def date_stat_hash(stat)
-        i = 0
         stat_hash = {}
-        keys = []
-        dates = []
-
-        while i < @days_previous
-          date = @start_date - i
-          datestr = date.strftime("%Y-%m-%d".freeze)
-          keys << "stat:#{stat}:#{datestr}"
-          dates << datestr
-          i += 1
-        end
+        dates = @start_date.downto(@start_date - @days_previous + 1).map { |date|
+          date.strftime("%Y-%m-%d")
+        }
 
-        Sidekiq.redis do |conn|
-          conn.mget(keys).each_with_index do |value, idx|
-            stat_hash[dates[idx]] = value ? value.to_i : 0
+        keys = dates.map { |datestr| "stat:#{stat}:#{datestr}" }
+
+        begin
+          Sidekiq.redis do |conn|
+            conn.mget(keys).each_with_index do |value, idx|
+              stat_hash[dates[idx]] = value ? value.to_i : 0
+            end
           end
+        rescue RedisConnection.adapter::CommandError
+          # mget will trigger a CROSSSLOT error when run against a Cluster
+          # TODO Someone want to add Cluster support?
         end
 
         stat_hash
@@ -181,9 +228,10 @@ module Sidekiq
   end
 
   ##
-  # Encapsulates a queue within Sidekiq.
+  # Represents a queue within Sidekiq.
   # Allows enumeration of all jobs within the queue
-  # and deletion of jobs.
+  # and deletion of jobs. NB: this queue data is real-time
+  # and is changing within Redis moment by moment.
   #
   #   queue = Sidekiq::Queue.new("mailer")
   #   queue.each do |job|
@@ -191,29 +239,34 @@ module Sidekiq
   #     job.args # => [1, 2, 3]
   #     job.delete if job.jid == 'abcdef1234567890'
   #   end
-  #
   class Queue
     include Enumerable
 
     ##
-    # Return all known queues within Redis.
+    # Fetch all known queues within Redis.
     #
+    # @return [Array<Sidekiq::Queue>]
     def self.all
-      Sidekiq.redis { |c| c.smembers('queues'.freeze) }.sort.map { |q| Sidekiq::Queue.new(q) }
+      Sidekiq.redis { |c| c.sscan_each("queues").to_a }.sort.map { |q| Sidekiq::Queue.new(q) }
     end
 
     attr_reader :name
 
-    def initialize(name="default")
-      @name = name
+    # @param name [String] the name of the queue
+    def initialize(name = "default")
+      @name = name.to_s
       @rname = "queue:#{name}"
     end
 
+    # The current size of the queue within Redis.
+    # This value is real-time and can change between calls.
+    #
+    # @return [Integer] the size
     def size
       Sidekiq.redis { |con| con.llen(@rname) }
     end
 
-    # Sidekiq Pro overrides this
+    # @return [Boolean] if the queue is currently paused
     def paused?
       false
     end
@@ -222,15 +275,15 @@ module Sidekiq
     # Calculates this queue's latency, the difference in seconds since the oldest
     # job in the queue was enqueued.
     #
-    # @return Float
+    # @return [Float] in seconds
     def latency
-      entry = Sidekiq.redis do |conn|
+      entry = Sidekiq.redis { |conn|
         conn.lrange(@rname, -1, -1)
-      end.first
+      }.first
       return 0 unless entry
       job = Sidekiq.load_json(entry)
       now = Time.now.to_f
-      thence = job['enqueued_at'] || now
+      thence = job["enqueued_at"] || now
       now - thence
     end
 
@@ -240,16 +293,16 @@ module Sidekiq
       page = 0
       page_size = 50
 
-      while true do
+      loop do
         range_start = page * page_size - deleted_size
-        range_end   = range_start + page_size - 1
-        entries = Sidekiq.redis do |conn|
+        range_end = range_start + page_size - 1
+        entries = Sidekiq.redis { |conn|
           conn.lrange @rname, range_start, range_end
-        end
+        }
         break if entries.empty?
         page += 1
         entries.each do |entry|
-          yield Job.new(entry, @name)
+          yield JobRecord.new(entry, @name)
         end
         deleted_size = initial_size - size
       end
@@ -258,150 +311,236 @@ module Sidekiq
     ##
     # Find the job with the given JID within this queue.
     #
-    # This is a slow, inefficient operation.  Do not use under
-    # normal conditions.  Sidekiq Pro contains a faster version.
+    # This is a *slow, inefficient* operation.  Do not use under
+    # normal conditions.
+    #
+    # @param jid [String] the job_id to look for
+    # @return [Sidekiq::JobRecord]
+    # @return [nil] if not found
     def find_job(jid)
       detect { |j| j.jid == jid }
     end
 
+    # delete all jobs within this queue
+    # @return [Boolean] true
     def clear
       Sidekiq.redis do |conn|
-        conn.multi do
-          conn.del(@rname)
-          conn.srem("queues".freeze, name)
+        conn.multi do |transaction|
+          transaction.unlink(@rname)
+          transaction.srem("queues", [name])
         end
       end
+      true
     end
     alias_method :💣, :clear
+
+    # :nodoc:
+    # @api private
+    def as_json(options = nil)
+      {name: name} # 5336
+    end
   end
 
   ##
-  # Encapsulates a pending job within a Sidekiq queue or
-  # sorted set.
+  # Represents a pending job within a Sidekiq queue.
   #
   # The job should be considered immutable but may be
-  # removed from the queue via Job#delete.
-  #
-  class Job
+  # removed from the queue via JobRecord#delete.
+  class JobRecord
+    # the parsed Hash of job data
+    # @!attribute [r] Item
     attr_reader :item
+    # the underlying String in Redis
+    # @!attribute [r] Value
     attr_reader :value
-
-    def initialize(item, queue_name=nil)
+    # the queue associated with this job
+    # @!attribute [r] Queue
+    attr_reader :queue
+
+    # :nodoc:
+    # @api private
+    def initialize(item, queue_name = nil)
+      @args = nil
       @value = item
-      @item = item.is_a?(Hash) ? item : Sidekiq.load_json(item)
-      @queue = queue_name || @item['queue']
+      @item = item.is_a?(Hash) ? item : parse(item)
+      @queue = queue_name || @item["queue"]
     end
 
+    # :nodoc:
+    # @api private
+    def parse(item)
+      Sidekiq.load_json(item)
+    rescue JSON::ParserError
+      # If the job payload in Redis is invalid JSON, we'll load
+      # the item as an empty hash and store the invalid JSON as
+      # the job 'args' for display in the Web UI.
+      @invalid = true
+      @args = [item]
+      {}
+    end
+
+    # This is the job class which Sidekiq will execute. If using ActiveJob,
+    # this class will be the ActiveJob adapter class rather than a specific job.
     def klass
-      @item['class']
+      self["class"]
     end
 
     def display_class
       # Unwrap known wrappers so they show up in a human-friendly manner in the Web UI
-      @klass ||= case klass
-                 when /\ASidekiq::Extensions::Delayed/
-                   safe_load(args[0], klass) do |target, method, _|
-                     "#{target}.#{method}"
-                   end
-                 when "ActiveJob::QueueAdapters::SidekiqAdapter::JobWrapper"
-                   job_class = @item['wrapped'] || args[0]
-                   if 'ActionMailer::DeliveryJob' == job_class
-                     # MailerClass#mailer_method
-                     args[0]['arguments'][0..1].join('#')
-                   else
-                    job_class
-                   end
-                 else
-                   klass
-                 end
+      @klass ||= self["display_class"] || begin
+        case klass
+        when /\ASidekiq::Extensions::Delayed/
+          safe_load(args[0], klass) do |target, method, _|
+            "#{target}.#{method}"
+          end
+        when "ActiveJob::QueueAdapters::SidekiqAdapter::JobWrapper"
+          job_class = @item["wrapped"] || args[0]
+          if job_class == "ActionMailer::DeliveryJob" || job_class == "ActionMailer::MailDeliveryJob"
+            # MailerClass#mailer_method
+            args[0]["arguments"][0..1].join("#")
+          else
+            job_class
+          end
+        else
+          klass
+        end
+      end
     end
 
     def display_args
       # Unwrap known wrappers so they show up in a human-friendly manner in the Web UI
-      @args ||= case klass
-                when /\ASidekiq::Extensions::Delayed/
-                  safe_load(args[0], args) do |_, _, arg|
-                    arg
-                  end
-                when "ActiveJob::QueueAdapters::SidekiqAdapter::JobWrapper"
-                  job_args = @item['wrapped'] ? args[0]["arguments"] : []
-                  if 'ActionMailer::DeliveryJob' == (@item['wrapped'] || args[0])
-                   # remove MailerClass, mailer_method and 'deliver_now'
-                   job_args.drop(3)
-                  else
-                   job_args
-                  end
-                else
-                  args
-                end
+      @display_args ||= case klass
+      when /\ASidekiq::Extensions::Delayed/
+        safe_load(args[0], args) do |_, _, arg, kwarg|
+          if !kwarg || kwarg.empty?
+            arg
+          else
+            [arg, kwarg]
+          end
+        end
+      when "ActiveJob::QueueAdapters::SidekiqAdapter::JobWrapper"
+        job_args = self["wrapped"] ? args[0]["arguments"] : []
+        if (self["wrapped"] || args[0]) == "ActionMailer::DeliveryJob"
+          # remove MailerClass, mailer_method and 'deliver_now'
+          job_args.drop(3)
+        elsif (self["wrapped"] || args[0]) == "ActionMailer::MailDeliveryJob"
+          # remove MailerClass, mailer_method and 'deliver_now'
+          job_args.drop(3).first["args"]
+        else
+          job_args
+        end
+      else
+        if self["encrypt"]
+          # no point in showing 150+ bytes of random garbage
+          args[-1] = "[encrypted data]"
+        end
+        args
+      end
     end
 
     def args
-      @item['args']
+      @args || @item["args"]
     end
 
     def jid
-      @item['jid']
+      self["jid"]
     end
 
     def enqueued_at
-      @item['enqueued_at'] ? Time.at(@item['enqueued_at']).utc : nil
+      self["enqueued_at"] ? Time.at(self["enqueued_at"]).utc : nil
     end
 
     def created_at
-      Time.at(@item['created_at'] || @item['enqueued_at'] || 0).utc
+      Time.at(self["created_at"] || self["enqueued_at"] || 0).utc
+    end
+
+    def tags
+      self["tags"] || []
     end
 
-    def queue
-      @queue
+    def error_backtrace
+      # Cache nil values
+      if defined?(@error_backtrace)
+        @error_backtrace
+      else
+        value = self["error_backtrace"]
+        @error_backtrace = value && uncompress_backtrace(value)
+      end
     end
 
     def latency
       now = Time.now.to_f
-      now - (@item['enqueued_at'] || @item['created_at'] || now)
+      now - (@item["enqueued_at"] || @item["created_at"] || now)
     end
 
-    ##
-    # Remove this job from the queue.
+    # Remove this job from the queue
     def delete
-      count = Sidekiq.redis do |conn|
+      count = Sidekiq.redis { |conn|
         conn.lrem("queue:#{@queue}", 1, @value)
-      end
+      }
       count != 0
     end
 
+    # Access arbitrary attributes within the job hash
     def [](name)
-      @item[name]
+      # nil will happen if the JSON fails to parse.
+      # We don't guarantee Sidekiq will work with bad job JSON but we should
+      # make a best effort to minimize the damage.
+      @item ? @item[name] : nil
     end
 
     private
 
     def safe_load(content, default)
-      begin
-        yield(*YAML.load(content))
-      rescue => ex
-        # #1761 in dev mode, it's possible to have jobs enqueued which haven't been loaded into
-        # memory yet so the YAML can't be loaded.
-        Sidekiq.logger.warn "Unable to load YAML: #{ex.message}" unless Sidekiq.options[:environment] == 'development'
-        default
+      yield(*YAML.load(content))
+    rescue => ex
+      # #1761 in dev mode, it's possible to have jobs enqueued which haven't been loaded into
+      # memory yet so the YAML can't be loaded.
+      # TODO is this still necessary? Zeitwerk reloader should handle?
+      Sidekiq.logger.warn "Unable to load YAML: #{ex.message}" unless Sidekiq.options[:environment] == "development"
+      default
+    end
+
+    def uncompress_backtrace(backtrace)
+      if backtrace.is_a?(Array)
+        # Handle old jobs with raw Array backtrace format
+        backtrace
+      else
+        decoded = Base64.decode64(backtrace)
+        uncompressed = Zlib::Inflate.inflate(decoded)
+        begin
+          Sidekiq.load_json(uncompressed)
+        rescue
+          # Handle old jobs with marshalled backtrace format
+          # TODO Remove in 7.x
+          Marshal.load(uncompressed)
+        end
       end
     end
   end
 
-  class SortedEntry < Job
+  # Represents a job within a Redis sorted set where the score
+  # represents a timestamp associated with the job. This timestamp
+  # could be the scheduled time for it to run (e.g. scheduled set),
+  # or the expiration date after which the entry should be deleted (e.g. dead set).
+  class SortedEntry < JobRecord
     attr_reader :score
     attr_reader :parent
 
+    # :nodoc:
+    # @api private
     def initialize(parent, score, item)
       super(item)
-      @score = score
+      @score = Float(score)
       @parent = parent
     end
 
+    # The timestamp associated with this entry
     def at
       Time.at(score).utc
     end
 
+    # remove this entry from the sorted set
     def delete
       if @value
         @parent.delete_by_value(@parent.name, @value)
@@ -410,11 +549,17 @@ module Sidekiq
       end
     end
 
+    # Change the scheduled time for this job.
+    #
+    # @param at [Time] the new timestamp for this job
     def reschedule(at)
-      delete
-      @parent.schedule(at, item)
+      Sidekiq.redis do |conn|
+        conn.zincrby(@parent.name, at.to_f - @score, Sidekiq.dump_json(@item))
+      end
     end
 
+    # Enqueue this job from the scheduled or dead set so it will
+    # be executed at some point in the near future.
     def add_to_queue
       remove_job do |message|
         msg = Sidekiq.load_json(message)
@@ -422,98 +567,127 @@ module Sidekiq
       end
     end
 
+    # enqueue this job from the retry set so it will be executed
+    # at some point in the near future.
     def retry
       remove_job do |message|
         msg = Sidekiq.load_json(message)
-        msg['retry_count'] -= 1 if msg['retry_count']
+        msg["retry_count"] -= 1 if msg["retry_count"]
         Sidekiq::Client.push(msg)
       end
     end
 
-    ##
-    # Place job in the dead set
+    # Move this job from its current set into the Dead set.
     def kill
       remove_job do |message|
-        now = Time.now.to_f
-        Sidekiq.redis do |conn|
-          conn.multi do
-            conn.zadd('dead', now, message)
-            conn.zremrangebyscore('dead', '-inf', now - DeadSet.timeout)
-            conn.zremrangebyrank('dead', 0, - DeadSet.max_jobs)
-          end
-        end
+        DeadSet.new.kill(message)
       end
     end
 
     def error?
-      !!item['error_class']
+      !!item["error_class"]
     end
 
     private
 
     def remove_job
       Sidekiq.redis do |conn|
-        results = conn.multi do
-          conn.zrangebyscore(parent.name, score, score)
-          conn.zremrangebyscore(parent.name, score, score)
-        end.first
+        results = conn.multi { |transaction|
+          transaction.zrangebyscore(parent.name, score, score)
+          transaction.zremrangebyscore(parent.name, score, score)
+        }.first
 
         if results.size == 1
           yield results.first
         else
           # multiple jobs with the same score
           # find the one with the right JID and push it
-          hash = results.group_by do |message|
+          matched, nonmatched = results.partition { |message|
             if message.index(jid)
               msg = Sidekiq.load_json(message)
-              msg['jid'] == jid
+              msg["jid"] == jid
             else
               false
             end
-          end
+          }
 
-          msg = hash.fetch(true, []).first
+          msg = matched.first
           yield msg if msg
 
           # push the rest back onto the sorted set
-          conn.multi do
-            hash.fetch(false, []).each do |message|
-              conn.zadd(parent.name, score.to_f.to_s, message)
+          conn.multi do |transaction|
+            nonmatched.each do |message|
+              transaction.zadd(parent.name, score.to_f.to_s, message)
             end
           end
         end
       end
     end
-
   end
 
+  # Base class for all sorted sets within Sidekiq.
   class SortedSet
     include Enumerable
 
+    # Redis key of the set
+    # @!attribute [r] Name
     attr_reader :name
 
+    # :nodoc:
+    # @api private
     def initialize(name)
       @name = name
       @_size = size
     end
 
+    # real-time size of the set, will change
     def size
       Sidekiq.redis { |c| c.zcard(name) }
     end
 
+    # Scan through each element of the sorted set, yielding each to the supplied block.
+    # Please see Redis's <a href="https://redis.io/commands/scan/">SCAN documentation</a> for implementation details.
+    #
+    # @param match [String] a snippet or regexp to filter matches.
+    # @param count [Integer] number of elements to retrieve at a time, default 100
+    # @yieldparam [Sidekiq::SortedEntry] each entry
+    def scan(match, count = 100)
+      return to_enum(:scan, match, count) unless block_given?
+
+      match = "*#{match}*" unless match.include?("*")
+      Sidekiq.redis do |conn|
+        conn.zscan_each(name, match: match, count: count) do |entry, score|
+          yield SortedEntry.new(self, score, entry)
+        end
+      end
+    end
+
+    # @return [Boolean] always true
     def clear
       Sidekiq.redis do |conn|
-        conn.del(name)
+        conn.unlink(name)
       end
+      true
     end
     alias_method :💣, :clear
+
+    # :nodoc:
+    # @api private
+    def as_json(options = nil)
+      {name: name} # 5336
+    end
   end
 
+  # Base class for all sorted sets which contain jobs, e.g. scheduled, retry and dead.
+  # Sidekiq Pro and Enterprise add additional sorted sets which do not contain job data,
+  # e.g. Batches.
   class JobSet < SortedSet
-
-    def schedule(timestamp, message)
+    # Add a job with the associated timestamp to this set.
+    # @param timestamp [Time] the score for the job
+    # @param job [Hash] the job data
+    def schedule(timestamp, job)
       Sidekiq.redis do |conn|
-        conn.zadd(name, timestamp.to_f.to_s, Sidekiq.dump_json(message))
+        conn.zadd(name, timestamp.to_f.to_s, Sidekiq.dump_json(job))
       end
     end
 
@@ -523,46 +697,66 @@ module Sidekiq
       page = -1
       page_size = 50
 
-      while true do
+      loop do
         range_start = page * page_size + offset_size
-        range_end   = range_start + page_size - 1
-        elements = Sidekiq.redis do |conn|
-          conn.zrange name, range_start, range_end, with_scores: true
-        end
+        range_end = range_start + page_size - 1
+        elements = Sidekiq.redis { |conn|
+          conn.zrange name, range_start, range_end, withscores: true
+        }
         break if elements.empty?
         page -= 1
-        elements.each do |element, score|
+        elements.reverse_each do |element, score|
           yield SortedEntry.new(self, score, element)
         end
         offset_size = initial_size - @_size
       end
     end
 
+    ##
+    # Fetch jobs that match a given time or Range. Job ID is an
+    # optional second argument.
+    #
+    # @param score [Time,Range] a specific timestamp or range
+    # @param jid [String, optional] find a specific JID within the score
+    # @return [Array<SortedEntry>] any results found, can be empty
     def fetch(score, jid = nil)
-      elements = Sidekiq.redis do |conn|
-        conn.zrangebyscore(name, score, score)
-      end
-
-      elements.inject([]) do |result, element|
-        entry = SortedEntry.new(self, score, element)
-        if jid
-          result << entry if entry.jid == jid
+      begin_score, end_score =
+        if score.is_a?(Range)
+          [score.first, score.last]
         else
-          result << entry
+          [score, score]
         end
-        result
+
+      elements = Sidekiq.redis { |conn|
+        conn.zrangebyscore(name, begin_score, end_score, withscores: true)
+      }
+
+      elements.each_with_object([]) do |element, result|
+        data, job_score = element
+        entry = SortedEntry.new(self, job_score, data)
+        result << entry if jid.nil? || entry.jid == jid
       end
     end
 
     ##
     # Find the job with the given JID within this sorted set.
+    # *This is a slow O(n) operation*.  Do not use for app logic.
     #
-    # This is a slow, inefficient operation.  Do not use under
-    # normal conditions.  Sidekiq Pro contains a faster version.
+    # @param jid [String] the job identifier
+    # @return [SortedEntry] the record or nil
     def find_job(jid)
-      self.detect { |j| j.jid == jid }
+      Sidekiq.redis do |conn|
+        conn.zscan_each(name, match: "*#{jid}*", count: 100) do |entry, score|
+          job = JSON.parse(entry)
+          matched = job["jid"] == jid
+          return SortedEntry.new(self, score, entry) if matched
+        end
+      end
+      nil
     end
 
+    # :nodoc:
+    # @api private
     def delete_by_value(name, value)
       Sidekiq.redis do |conn|
         ret = conn.zrem(name, value)
@@ -571,17 +765,20 @@ module Sidekiq
       end
     end
 
+    # :nodoc:
+    # @api private
     def delete_by_jid(score, jid)
       Sidekiq.redis do |conn|
         elements = conn.zrangebyscore(name, score, score)
         elements.each do |element|
-          message = Sidekiq.load_json(element)
-          if message["jid"] == jid
-            ret = conn.zrem(name, element)
-            @_size -= 1 if ret
-            break ret
+          if element.index(jid)
+            message = Sidekiq.load_json(element)
+            if message["jid"] == jid
+              ret = conn.zrem(name, element)
+              @_size -= 1 if ret
+              break ret
+            end
           end
-          false
         end
       end
     end
@@ -590,10 +787,10 @@ module Sidekiq
   end
 
   ##
-  # Allows enumeration of scheduled jobs within Sidekiq.
+  # The set of scheduled jobs within Sidekiq.
   # Based on this, you can search/filter for jobs.  Here's an
-  # example where I'm selecting all jobs of a certain type
-  # and deleting them from the schedule queue.
+  # example where I'm selecting jobs based on some complex logic
+  # and deleting them from the scheduled set.
   #
   #   r = Sidekiq::ScheduledSet.new
   #   r.select do |scheduled|
@@ -603,12 +800,12 @@ module Sidekiq
   #   end.map(&:delete)
   class ScheduledSet < JobSet
     def initialize
-      super 'schedule'
+      super "schedule"
     end
   end
 
   ##
-  # Allows enumeration of retries within Sidekiq.
+  # The set of retries within Sidekiq.
   # Based on this, you can search/filter for jobs.  Here's an
   # example where I'm selecting all jobs of a certain type
   # and deleting them from the retry queue.
@@ -621,105 +818,179 @@ module Sidekiq
   #   end.map(&:delete)
   class RetrySet < JobSet
     def initialize
-      super 'retry'
+      super "retry"
     end
 
+    # Enqueues all jobs pending within the retry set.
     def retry_all
-      while size > 0
-        each(&:retry)
-      end
+      each(&:retry) while size > 0
+    end
+
+    # Kills all jobs pending within the retry set.
+    def kill_all
+      each(&:kill) while size > 0
     end
   end
 
   ##
-  # Allows enumeration of dead jobs within Sidekiq.
+  # The set of dead jobs within Sidekiq. Dead jobs have failed all of
+  # their retries and are helding in this set pending some sort of manual
+  # fix. They will be removed after 6 months (dead_timeout) if not.
   #
   class DeadSet < JobSet
     def initialize
-      super 'dead'
+      super "dead"
     end
 
-    def retry_all
-      while size > 0
-        each(&:retry)
+    # Add the given job to the Dead set.
+    # @param message [String] the job data as JSON
+    def kill(message, opts = {})
+      now = Time.now.to_f
+      Sidekiq.redis do |conn|
+        conn.multi do |transaction|
+          transaction.zadd(name, now.to_s, message)
+          transaction.zremrangebyscore(name, "-inf", now - self.class.timeout)
+          transaction.zremrangebyrank(name, 0, - self.class.max_jobs)
+        end
       end
+
+      if opts[:notify_failure] != false
+        job = Sidekiq.load_json(message)
+        r = RuntimeError.new("Job killed by API")
+        r.set_backtrace(caller)
+        Sidekiq.death_handlers.each do |handle|
+          handle.call(job, r)
+        end
+      end
+      true
     end
 
+    # Enqueue all dead jobs
+    def retry_all
+      each(&:retry) while size > 0
+    end
+
+    # The maximum size of the Dead set. Older entries will be trimmed
+    # to stay within this limit. Default value is 10,000.
     def self.max_jobs
-      Sidekiq.options[:dead_max_jobs]
+      Sidekiq[:dead_max_jobs]
     end
 
+    # The time limit for entries within the Dead set. Older entries will be thrown away.
+    # Default value is six months.
     def self.timeout
-      Sidekiq.options[:dead_timeout_in_seconds]
+      Sidekiq[:dead_timeout_in_seconds]
     end
   end
 
   ##
   # Enumerates the set of Sidekiq processes which are actively working
-  # right now.  Each process send a heartbeat to Redis every 5 seconds
+  # right now.  Each process sends a heartbeat to Redis every 5 seconds
   # so this set should be relatively accurate, barring network partitions.
   #
-  # Yields a Sidekiq::Process.
+  # @yieldparam [Sidekiq::Process]
   #
   class ProcessSet
     include Enumerable
 
-    def initialize(clean_plz=true)
-      self.class.cleanup if clean_plz
+    # :nodoc:
+    # @api private
+    def initialize(clean_plz = true)
+      cleanup if clean_plz
     end
 
     # Cleans up dead processes recorded in Redis.
     # Returns the number of processes cleaned.
-    def self.cleanup
+    # :nodoc:
+    # @api private
+    def cleanup
+      # dont run cleanup more than once per minute
+      return 0 unless Sidekiq.redis { |conn| conn.set("process_cleanup", "1", nx: true, ex: 60) }
+
       count = 0
       Sidekiq.redis do |conn|
-        procs = conn.smembers('processes').sort
-        heartbeats = conn.pipelined do
+        procs = conn.sscan_each("processes").to_a
+        heartbeats = conn.pipelined { |pipeline|
           procs.each do |key|
-            conn.hget(key, 'info')
+            pipeline.hget(key, "info")
           end
-        end
+        }
 
         # the hash named key has an expiry of 60 seconds.
         # if it's not found, that means the process has not reported
         # in to Redis and probably died.
-        to_prune = []
-        heartbeats.each_with_index do |beat, i|
-          to_prune << procs[i] if beat.nil?
-        end
-        count = conn.srem('processes', to_prune) unless to_prune.empty?
+        to_prune = procs.select.with_index { |proc, i|
+          heartbeats[i].nil?
+        }
+        count = conn.srem("processes", to_prune) unless to_prune.empty?
       end
       count
     end
 
     def each
-      procs = Sidekiq.redis { |conn| conn.smembers('processes') }.sort
+      result = Sidekiq.redis { |conn|
+        procs = conn.sscan_each("processes").to_a.sort
 
-      Sidekiq.redis do |conn|
         # We're making a tradeoff here between consuming more memory instead of
         # making more roundtrips to Redis, but if you have hundreds or thousands of workers,
         # you'll be happier this way
-        result = conn.pipelined do
+        conn.pipelined do |pipeline|
           procs.each do |key|
-            conn.hmget(key, 'info', 'busy', 'beat', 'quiet')
+            pipeline.hmget(key, "info", "busy", "beat", "quiet", "rss", "rtt_us")
           end
         end
+      }
 
-        result.each do |info, busy, at_s, quiet|
-          hash = Sidekiq.load_json(info)
-          yield Process.new(hash.merge('busy' => busy.to_i, 'beat' => at_s.to_f, 'quiet' => quiet))
-        end
+      result.each do |info, busy, at_s, quiet, rss, rtt|
+        # If a process is stopped between when we query Redis for `procs` and
+        # when we query for `result`, we will have an item in `result` that is
+        # composed of `nil` values.
+        next if info.nil?
+
+        hash = Sidekiq.load_json(info)
+        yield Process.new(hash.merge("busy" => busy.to_i,
+          "beat" => at_s.to_f,
+          "quiet" => quiet,
+          "rss" => rss.to_i,
+          "rtt_us" => rtt.to_i))
       end
-
-      nil
     end
 
     # This method is not guaranteed accurate since it does not prune the set
     # based on current heartbeat.  #each does that and ensures the set only
     # contains Sidekiq processes which have sent a heartbeat within the last
     # 60 seconds.
+    # @return [Integer] current number of registered Sidekiq processes
     def size
-      Sidekiq.redis { |conn| conn.scard('processes') }
+      Sidekiq.redis { |conn| conn.scard("processes") }
+    end
+
+    # Total number of threads available to execute jobs.
+    # For Sidekiq Enterprise customers this number (in production) must be
+    # less than or equal to your licensed concurrency.
+    # @return [Integer] the sum of process concurrency
+    def total_concurrency
+      sum { |x| x["concurrency"].to_i }
+    end
+
+    # @return [Integer] total amount of RSS memory consumed by Sidekiq processes
+    def total_rss_in_kb
+      sum { |x| x["rss"].to_i }
+    end
+    alias_method :total_rss, :total_rss_in_kb
+
+    # Returns the identity of the current cluster leader or "" if no leader.
+    # This is a Sidekiq Enterprise feature, will always return "" in Sidekiq
+    # or Sidekiq Pro.
+    # @return [String] Identity of cluster leader
+    # @return [String] empty string if no leader
+    def leader
+      @leader ||= begin
+        x = Sidekiq.redis { |c| c.get("dear-leader") }
+        # need a non-falsy value so we can memoize
+        x ||= ""
+        x
+      end
     end
   end
 
@@ -739,36 +1010,59 @@ module Sidekiq
   #   'identity' => <unique string identifying the process>,
   # }
   class Process
+    # :nodoc:
+    # @api private
     def initialize(hash)
       @attribs = hash
     end
 
     def tag
-      self['tag']
+      self["tag"]
     end
 
     def labels
-      Array(self['labels'])
+      Array(self["labels"])
     end
 
     def [](key)
       @attribs[key]
     end
 
+    def identity
+      self["identity"]
+    end
+
+    def queues
+      self["queues"]
+    end
+
+    # Signal this process to stop processing new jobs.
+    # It will continue to execute jobs it has already fetched.
+    # This method is *asynchronous* and it can take 5-10
+    # seconds for the process to quiet.
     def quiet!
-      signal('USR1')
+      signal("TSTP")
     end
 
+    # Signal this process to shutdown.
+    # It will shutdown within its configured :timeout value, default 25 seconds.
+    # This method is *asynchronous* and it can take 5-10
+    # seconds for the process to start shutting down.
     def stop!
-      signal('TERM')
+      signal("TERM")
     end
 
+    # Signal this process to log backtraces for all threads.
+    # Useful if you have a frozen or deadlocked process which is
+    # still sending a heartbeat.
+    # This method is *asynchronous* and it can take 5-10 seconds.
     def dump_threads
-      signal('TTIN')
+      signal("TTIN")
     end
 
+    # @return [Boolean] true if this process is quiet or shutting down
     def stopping?
-      self['quiet'] == 'true'
+      self["quiet"] == "true"
     end
 
     private
@@ -776,21 +1070,17 @@ module Sidekiq
     def signal(sig)
       key = "#{identity}-signals"
       Sidekiq.redis do |c|
-        c.multi do
-          c.lpush(key, sig)
-          c.expire(key, 60)
+        c.multi do |transaction|
+          transaction.lpush(key, sig)
+          transaction.expire(key, 60)
         end
       end
     end
-
-    def identity
-      self['identity']
-    end
   end
 
   ##
-  # A worker is a thread that is currently processing a job.
-  # Programmatic access to the current active worker set.
+  # The WorkSet stores the work being done by this Sidekiq cluster.
+  # It tracks the process and thread working on each job.
   #
   # WARNING WARNING WARNING
   #
@@ -798,33 +1088,40 @@ module Sidekiq
   # If you call #size => 5 and then expect #each to be
   # called 5 times, you're going to have a bad time.
   #
-  #    workers = Sidekiq::Workers.new
-  #    workers.size => 2
-  #    workers.each do |process_id, thread_id, work|
+  #    works = Sidekiq::WorkSet.new
+  #    works.size => 2
+  #    works.each do |process_id, thread_id, work|
   #      # process_id is a unique identifier per Sidekiq process
   #      # thread_id is a unique identifier per thread
   #      # work is a Hash which looks like:
-  #      # { 'queue' => name, 'run_at' => timestamp, 'payload' => msg }
+  #      # { 'queue' => name, 'run_at' => timestamp, 'payload' => job_hash }
   #      # run_at is an epoch Integer.
   #    end
   #
-  class Workers
+  class WorkSet
     include Enumerable
 
-    def each
+    def each(&block)
+      results = []
       Sidekiq.redis do |conn|
-        procs = conn.smembers('processes')
+        procs = conn.sscan_each("processes").to_a
         procs.sort.each do |key|
-          valid, workers = conn.pipelined do
-            conn.exists(key)
-            conn.hgetall("#{key}:workers")
-          end
+          valid, workers = conn.pipelined { |pipeline|
+            pipeline.exists?(key)
+            pipeline.hgetall("#{key}:work")
+          }
           next unless valid
           workers.each_pair do |tid, json|
-            yield key, tid, Sidekiq.load_json(json)
+            hsh = Sidekiq.load_json(json)
+            p = hsh["payload"]
+            # avoid breaking API, this is a side effect of the JSON optimization in #4316
+            hsh["payload"] = Sidekiq.load_json(p) if p.is_a?(String)
+            results << [key, tid, hsh]
           end
         end
       end
+
+      results.sort_by { |(_, _, hsh)| hsh["run_at"] }.each(&block)
     end
 
     # Note that #size is only as accurate as Sidekiq's heartbeat,
@@ -835,18 +1132,21 @@ module Sidekiq
     # which can easily get out of sync with crashy processes.
     def size
       Sidekiq.redis do |conn|
-        procs = conn.smembers('processes')
+        procs = conn.sscan_each("processes").to_a
         if procs.empty?
           0
         else
-          conn.pipelined do
+          conn.pipelined { |pipeline|
             procs.each do |key|
-              conn.hget(key, 'busy')
+              pipeline.hget(key, "busy")
             end
-          end.map(&:to_i).inject(:+)
+          }.sum(&:to_i)
         end
       end
     end
   end
-
+  # Since "worker" is a nebulous term, we've deprecated the use of this class name.
+  # Is "worker" a process, a type of job, a thread? Undefined!
+  # WorkSet better describes the data.
+  Workers = WorkSet
 end