sidekiq 4.2.4 → 6.4.0

data/lib/sidekiq/rails.rb

@@ -1,81 +1,11 @@
 # frozen_string_literal: true
-module Sidekiq
-  def self.hook_rails!
-    return if defined?(@delay_removed)
-
-    ActiveSupport.on_load(:active_record) do
-      include Sidekiq::Extensions::ActiveRecord
-    end
 
-    ActiveSupport.on_load(:action_mailer) do
-      extend Sidekiq::Extensions::ActionMailer
-    end
-
-    Module.__send__(:include, Sidekiq::Extensions::Klass)
-  end
-
-  # Removes the generic aliases which MAY clash with names of already
-  #  created methods by other applications. The methods `sidekiq_delay`,
-  #  `sidekiq_delay_for` and `sidekiq_delay_until` can be used instead.
-  def self.remove_delay!
-    @delay_removed = true
-
-    [Extensions::ActiveRecord,
-     Extensions::ActionMailer,
-     Extensions::Klass].each do |mod|
-      mod.module_eval do
-        remove_method :delay if respond_to?(:delay)
-        remove_method :delay_for if respond_to?(:delay_for)
-        remove_method :delay_until if respond_to?(:delay_until)
-      end
-    end
-  end
+require "sidekiq/worker"
 
+module Sidekiq
   class Rails < ::Rails::Engine
-    # We need to setup this up before any application configuration which might
-    # change Sidekiq middleware.
-    #
-    # This hook happens after `Rails::Application` is inherited within
-    # config/application.rb and before config is touched, usually within the
-    # class block. Definitely before config/environments/*.rb and
-    # config/initializers/*.rb.
-    config.before_configuration do
-      if ::Rails::VERSION::MAJOR < 5 && defined?(::ActiveRecord)
-        Sidekiq.server_middleware do |chain|
-          require 'sidekiq/middleware/server/active_record'
-          chain.add Sidekiq::Middleware::Server::ActiveRecord
-        end
-      end
-    end
-
-    initializer 'sidekiq' do
-      Sidekiq.hook_rails!
-    end
-
-    # We have to add the reloader after initialize to see if cache_classes has
-    # been turned on.
-    #
-    # This hook happens after all initialziers are run, just before returning
-    # from config/environment.rb back to sidekiq/cli.rb.
-    config.after_initialize do
-      if ::Rails::VERSION::MAJOR >= 5
-        # The reloader also takes care of ActiveRecord but is incompatible with
-        # the ActiveRecord middleware so make sure it's not in the chain already.
-        if defined?(Sidekiq::Middleware::Server::ActiveRecord) && Sidekiq.server_middleware.exists?(Sidekiq::Middleware::Server::ActiveRecord)
-          raise ArgumentError, "You are using the Sidekiq ActiveRecord middleware and the new Rails 5 reloader which are incompatible. Please remove the ActiveRecord middleware from your Sidekiq middleware configuration."
-        elsif ::Rails.application.config.cache_classes
-          # The reloader API has proven to be troublesome under load in production.
-          # We won't use it at all when classes are cached, see #3154
-          Sidekiq.logger.debug { "Autoload disabled in #{::Rails.env}, Sidekiq will not reload changed classes" }
-        else
-          Sidekiq.options[:reloader] = Sidekiq::Rails::Reloader.new
-        end
-      end
-    end
-
     class Reloader
       def initialize(app = ::Rails.application)
-        Sidekiq.logger.debug "Enabling Rails 5+ live code reloading, so hot!" unless app.config.cache_classes
         @app = app
       end
 
@@ -89,5 +19,43 @@ module Sidekiq
         "#<Sidekiq::Rails::Reloader @app=#{@app.class.name}>"
       end
     end
-  end if defined?(::Rails)
+
+    # By including the Options module, we allow AJs to directly control sidekiq features
+    # via the *sidekiq_options* class method and, for instance, not use AJ's retry system.
+    # AJ retries don't show up in the Sidekiq UI Retries tab, save any error data, can't be
+    # manually retried, don't automatically die, etc.
+    #
+    #   class SomeJob < ActiveJob::Base
+    #     queue_as :default
+    #     sidekiq_options retry: 3, backtrace: 10
+    #     def perform
+    #     end
+    #   end
+    initializer "sidekiq.active_job_integration" do
+      ActiveSupport.on_load(:active_job) do
+        include ::Sidekiq::Worker::Options unless respond_to?(:sidekiq_options)
+      end
+    end
+
+    initializer "sidekiq.rails_logger" do
+      Sidekiq.configure_server do |_|
+        # This is the integration code necessary so that if code uses `Rails.logger.info "Hello"`,
+        # it will appear in the Sidekiq console with all of the job context. See #5021 and
+        # https://github.com/rails/rails/blob/b5f2b550f69a99336482739000c58e4e04e033aa/railties/lib/rails/commands/server/server_command.rb#L82-L84
+        unless ::Rails.logger == ::Sidekiq.logger || ::ActiveSupport::Logger.logger_outputs_to?(::Rails.logger, $stdout)
+          ::Rails.logger.extend(::ActiveSupport::Logger.broadcast(::Sidekiq.logger))
+        end
+      end
+    end
+
+    # This hook happens after all initializers are run, just before returning
+    # from config/environment.rb back to sidekiq/cli.rb.
+    #
+    # None of this matters on the client-side, only within the Sidekiq process itself.
+    config.after_initialize do
+      Sidekiq.configure_server do |_|
+        Sidekiq.options[:reloader] = Sidekiq::Rails::Reloader.new
+      end
+    end
+  end
 end

data/lib/sidekiq/redis_connection.rb

@@ -1,26 +1,38 @@
 # frozen_string_literal: true
-require 'connection_pool'
-require 'redis'
-require 'uri'
+
+require "connection_pool"
+require "redis"
+require "uri"
 
 module Sidekiq
   class RedisConnection
     class << self
+      def create(options = {})
+        symbolized_options = options.transform_keys(&:to_sym)
 
-      def create(options={})
-        options = options.symbolize_keys
-
-        options[:url] ||= determine_redis_provider
+        if !symbolized_options[:url] && (u = determine_redis_provider)
+          symbolized_options[:url] = u
+        end
 
-        size = options[:size] || (Sidekiq.server? ? (Sidekiq.options[:concurrency] + 5) : 5)
+        size = if symbolized_options[:size]
+          symbolized_options[:size]
+        elsif Sidekiq.server?
+          # Give ourselves plenty of connections.  pool is lazy
+          # so we won't create them until we need them.
+          Sidekiq.options[:concurrency] + 5
+        elsif ENV["RAILS_MAX_THREADS"]
+          Integer(ENV["RAILS_MAX_THREADS"])
+        else
+          5
+        end
 
         verify_sizing(size, Sidekiq.options[:concurrency]) if Sidekiq.server?
 
-        pool_timeout = options[:pool_timeout] || 1
-        log_info(options)
+        pool_timeout = symbolized_options[:pool_timeout] || 1
+        log_info(symbolized_options)
 
-        ConnectionPool.new(:timeout => pool_timeout, :size => size) do
-          build_client(options)
+        ConnectionPool.new(timeout: pool_timeout, size: size) do
+          build_client(symbolized_options)
         end
       end
 
@@ -35,7 +47,7 @@ module Sidekiq
       #   - enterprise's leader election
       #   - enterprise's cron support
       def verify_sizing(size, concurrency)
-        raise ArgumentError, "Your Redis connection pool is too small for Sidekiq to work. Your pool has #{size} connections but really needs to have at least #{concurrency + 2}" if size <= concurrency
+        raise ArgumentError, "Your Redis connection pool is too small for Sidekiq to work. Your pool has #{size} connections but must have at least #{concurrency + 2}" if size < (concurrency + 2)
       end
 
       def build_client(options)
@@ -44,8 +56,8 @@ module Sidekiq
         client = Redis.new client_opts(options)
         if namespace
           begin
-            require 'redis/namespace'
-            Redis::Namespace.new(namespace, :redis => client)
+            require "redis/namespace"
+            Redis::Namespace.new(namespace, redis: client)
           rescue LoadError
             Sidekiq.logger.error("Your Redis configuration uses the namespace '#{namespace}' but the redis-namespace gem is not included in the Gemfile." \
                                  "Add the gem to your Gemfile to continue using a namespace. Otherwise, remove the namespace parameter.")
@@ -67,15 +79,26 @@ module Sidekiq
           opts.delete(:network_timeout)
         end
 
-        opts[:driver] = opts[:driver] || 'ruby'
+        opts[:driver] ||= Redis::Connection.drivers.last || "ruby"
+
+        # Issue #3303, redis-rb will silently retry an operation.
+        # This can lead to duplicate jobs if Sidekiq::Client's LPUSH
+        # is performed twice but I believe this is much, much rarer
+        # than the reconnect silently fixing a problem; we keep it
+        # on by default.
+        opts[:reconnect_attempts] ||= 1
 
         opts
       end
 
       def log_info(options)
-        # Don't log Redis AUTH password
         redacted = "REDACTED"
-        scrubbed_options = options.dup
+
+        # Deep clone so we can muck with these options all we want and exclude
+        # params from dump-and-load that may contain objects that Marshal is
+        # unable to safely dump.
+        keys = options.keys - [:logger, :ssl_params]
+        scrubbed_options = Marshal.load(Marshal.dump(options.slice(*keys)))
         if scrubbed_options[:url] && (uri = URI.parse(scrubbed_options[:url])) && uri.password
           uri.password = redacted
           scrubbed_options[:url] = uri.to_s
@@ -83,6 +106,9 @@ module Sidekiq
         if scrubbed_options[:password]
           scrubbed_options[:password] = redacted
         end
+        scrubbed_options[:sentinels]&.each do |sentinel|
+          sentinel[:password] = redacted if sentinel[:password]
+        end
         if Sidekiq.server?
           Sidekiq.logger.info("Booting Sidekiq #{Sidekiq::VERSION} with redis options #{scrubbed_options}")
         else
@@ -91,9 +117,28 @@ module Sidekiq
       end
 
       def determine_redis_provider
-        ENV[ENV['REDIS_PROVIDER'] || 'REDIS_URL']
-      end
+        # If you have this in your environment:
+        # MY_REDIS_URL=redis://hostname.example.com:1238/4
+        # then set:
+        # REDIS_PROVIDER=MY_REDIS_URL
+        # and Sidekiq will find your custom URL variable with no custom
+        # initialization code at all.
+        #
+        p = ENV["REDIS_PROVIDER"]
+        if p && p =~ /:/
+          raise <<~EOM
+            REDIS_PROVIDER should be set to the name of the variable which contains the Redis URL, not a URL itself.
+            Platforms like Heroku will sell addons that publish a *_URL variable.  You need to tell Sidekiq with REDIS_PROVIDER, e.g.:
+
+            REDISTOGO_URL=redis://somehost.example.com:6379/4
+            REDIS_PROVIDER=REDISTOGO_URL
+          EOM
+        end
 
+        ENV[
+          p || "REDIS_URL"
+        ]
+      end
     end
   end
 end

data/lib/sidekiq/scheduled.rb

@@ -1,35 +1,64 @@
 # frozen_string_literal: true
-require 'sidekiq'
-require 'sidekiq/util'
-require 'sidekiq/api'
+
+require "sidekiq"
+require "sidekiq/util"
+require "sidekiq/api"
 
 module Sidekiq
   module Scheduled
-    SETS = %w(retry schedule)
+    SETS = %w[retry schedule]
 
     class Enq
-      def enqueue_jobs(now=Time.now.to_f.to_s, sorted_sets=SETS)
+      LUA_ZPOPBYSCORE = <<~LUA
+        local key, now = KEYS[1], ARGV[1]
+        local jobs = redis.call("zrangebyscore", key, "-inf", now, "limit", 0, 1)
+        if jobs[1] then
+          redis.call("zrem", key, jobs[1])
+          return jobs[1]
+        end
+      LUA
+
+      def initialize
+        @done = false
+        @lua_zpopbyscore_sha = nil
+      end
+
+      def enqueue_jobs(sorted_sets = SETS)
         # A job's "score" in Redis is the time at which it should be processed.
         # Just check Redis for the set of jobs with a timestamp before now.
         Sidekiq.redis do |conn|
           sorted_sets.each do |sorted_set|
-            # Get the next item in the queue if it's score (time to execute) is <= now.
+            # Get next item in the queue with score (time to execute) <= now.
             # We need to go through the list one at a time to reduce the risk of something
             # going wrong between the time jobs are popped from the scheduled queue and when
             # they are pushed onto a work queue and losing the jobs.
-            while job = conn.zrangebyscore(sorted_set, '-inf'.freeze, now, :limit => [0, 1]).first do
-
-              # Pop item off the queue and add it to the work queue. If the job can't be popped from
-              # the queue, it's because another process already popped it so we can move on to the
-              # next one.
-              if conn.zrem(sorted_set, job)
-                Sidekiq::Client.push(Sidekiq.load_json(job))
-                Sidekiq::Logging.logger.debug { "enqueued #{sorted_set}: #{job}" }
-              end
+            while !@done && (job = zpopbyscore(conn, keys: [sorted_set], argv: [Time.now.to_f.to_s]))
+              Sidekiq::Client.push(Sidekiq.load_json(job))
+              Sidekiq.logger.debug { "enqueued #{sorted_set}: #{job}" }
             end
           end
         end
       end
+
+      def terminate
+        @done = true
+      end
+
+      private
+
+      def zpopbyscore(conn, keys: nil, argv: nil)
+        if @lua_zpopbyscore_sha.nil?
+          raw_conn = conn.respond_to?(:redis) ? conn.redis : conn
+          @lua_zpopbyscore_sha = raw_conn.script(:load, LUA_ZPOPBYSCORE)
+        end
+
+        conn.evalsha(@lua_zpopbyscore_sha, keys: keys, argv: argv)
+      rescue Redis::CommandError => e
+        raise unless e.message.start_with?("NOSCRIPT")
+
+        @lua_zpopbyscore_sha = nil
+        retry
+      end
     end
 
     ##
@@ -47,11 +76,14 @@ module Sidekiq
         @sleeper = ConnectionPool::TimedStack.new
         @done = false
         @thread = nil
+        @count_calls = 0
       end
 
       # Shut down this instance, will pause until the thread is dead.
       def terminate
         @done = true
+        @enq.terminate if @enq.respond_to?(:terminate)
+
         if @thread
           t = @thread
           @thread = nil
@@ -61,28 +93,24 @@ module Sidekiq
       end
 
       def start
-        @thread ||= safe_thread("scheduler") do
+        @thread ||= safe_thread("scheduler") {
           initial_wait
 
-          while !@done
+          until @done
             enqueue
             wait
           end
           Sidekiq.logger.info("Scheduler exiting...")
-        end
+        }
       end
 
       def enqueue
-        begin
-          @enq.enqueue_jobs
-        rescue => ex
-          # Most likely a problem with redis networking.
-          # Punt and try again at the next interval
-          logger.error ex.message
-          ex.backtrace.each do |bt|
-            logger.error(bt)
-          end
-        end
+        @enq.enqueue_jobs
+      rescue => ex
+        # Most likely a problem with redis networking.
+        # Punt and try again at the next interval
+        logger.error ex.message
+        handle_exception(ex)
       end
 
       private
@@ -95,13 +123,38 @@ module Sidekiq
         # if poll_interval_average hasn't been calculated yet, we can
         # raise an error trying to reach Redis.
         logger.error ex.message
-        logger.error ex.backtrace.first
+        handle_exception(ex)
         sleep 5
       end
 
-      # Calculates a random interval that is ±50% the desired average.
       def random_poll_interval
-        poll_interval_average * rand + poll_interval_average.to_f / 2
+        # We want one Sidekiq process to schedule jobs every N seconds.  We have M processes
+        # and **don't** want to coordinate.
+        #
+        # So in N*M second timespan, we want each process to schedule once.  The basic loop is:
+        #
+        # * sleep a random amount within that N*M timespan
+        # * wake up and schedule
+        #
+        # We want to avoid one edge case: imagine a set of 2 processes, scheduling every 5 seconds,
+        # so N*M = 10.  Each process decides to randomly sleep 8 seconds, now we've failed to meet
+        # that 5 second average. Thankfully each schedule cycle will sleep randomly so the next
+        # iteration could see each process sleep for 1 second, undercutting our average.
+        #
+        # So below 10 processes, we special case and ensure the processes sleep closer to the average.
+        # In the example above, each process should schedule every 10 seconds on average. We special
+        # case smaller clusters to add 50% so they would sleep somewhere between 5 and 15 seconds.
+        # As we run more processes, the scheduling interval average will approach an even spread
+        # between 0 and poll interval so we don't need this artifical boost.
+        #
+        if process_count < 10
+          # For small clusters, calculate a random interval that is ±50% the desired average.
+          poll_interval_average * rand + poll_interval_average.to_f / 2
+        else
+          # With 10+ processes, we should have enough randomness to get decent polling
+          # across the entire timespan
+          poll_interval_average * rand
+        end
       end
 
       # We do our best to tune the poll interval to the size of the active Sidekiq
@@ -125,9 +178,18 @@ module Sidekiq
       # This minimizes a single point of failure by dispersing check-ins but without taxing
       # Redis if you run many Sidekiq processes.
       def scaled_poll_interval
-        pcount = Sidekiq::ProcessSet.new.size
+        process_count * Sidekiq.options[:average_scheduled_poll_interval]
+      end
+
+      def process_count
+        # The work buried within Sidekiq::ProcessSet#cleanup can be
+        # expensive at scale. Cut it down by 90% with this counter.
+        # NB: This method is only called by the scheduler thread so we
+        # don't need to worry about the thread safety of +=.
+        pcount = Sidekiq::ProcessSet.new(@count_calls % 10 == 0).size
         pcount = 1 if pcount == 0
-        pcount * Sidekiq.options[:average_scheduled_poll_interval]
+        @count_calls += 1
+        pcount
       end
 
       def initial_wait
@@ -141,7 +203,6 @@ module Sidekiq
         @sleeper.pop(total)
       rescue Timeout::Error
       end
-
     end
   end
 end

data/lib/sidekiq/sd_notify.rb

@@ -0,0 +1,149 @@
+# frozen_string_literal: true
+
+# The MIT License
+#
+# Copyright (c) 2017, 2018, 2019, 2020 Agis Anastasopoulos
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy of
+# this software and associated documentation files (the "Software"), to deal in
+# the Software without restriction, including without limitation the rights to
+# use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of
+# the Software, and to permit persons to whom the Software is furnished to do so,
+# subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in all
+# copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
+# FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
+# COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
+# IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+# This is a copy of https://github.com/agis/ruby-sdnotify as of commit a7d52ee
+# The only changes made was "rehoming" it within the Sidekiq module to avoid
+# namespace collisions and applying standard's code formatting style.
+
+require "socket"
+
+# SdNotify is a pure-Ruby implementation of sd_notify(3). It can be used to
+# notify systemd about state changes. Methods of this package are no-op on
+# non-systemd systems (eg. Darwin).
+#
+# The API maps closely to the original implementation of sd_notify(3),
+# therefore be sure to check the official man pages prior to using SdNotify.
+#
+# @see https://www.freedesktop.org/software/systemd/man/sd_notify.html
+module Sidekiq
+  module SdNotify
+    # Exception raised when there's an error writing to the notification socket
+    class NotifyError < RuntimeError; end
+
+    READY = "READY=1"
+    RELOADING = "RELOADING=1"
+    STOPPING = "STOPPING=1"
+    STATUS = "STATUS="
+    ERRNO = "ERRNO="
+    MAINPID = "MAINPID="
+    WATCHDOG = "WATCHDOG=1"
+    FDSTORE = "FDSTORE=1"
+
+    def self.ready(unset_env = false)
+      notify(READY, unset_env)
+    end
+
+    def self.reloading(unset_env = false)
+      notify(RELOADING, unset_env)
+    end
+
+    def self.stopping(unset_env = false)
+      notify(STOPPING, unset_env)
+    end
+
+    # @param status [String] a custom status string that describes the current
+    #   state of the service
+    def self.status(status, unset_env = false)
+      notify("#{STATUS}#{status}", unset_env)
+    end
+
+    # @param errno [Integer]
+    def self.errno(errno, unset_env = false)
+      notify("#{ERRNO}#{errno}", unset_env)
+    end
+
+    # @param pid [Integer]
+    def self.mainpid(pid, unset_env = false)
+      notify("#{MAINPID}#{pid}", unset_env)
+    end
+
+    def self.watchdog(unset_env = false)
+      notify(WATCHDOG, unset_env)
+    end
+
+    def self.fdstore(unset_env = false)
+      notify(FDSTORE, unset_env)
+    end
+
+    # @return [Boolean] true if the service manager expects watchdog keep-alive
+    #   notification messages to be sent from this process.
+    #
+    # If the $WATCHDOG_USEC environment variable is set,
+    # and the $WATCHDOG_PID variable is unset or set to the PID of the current
+    # process
+    #
+    # @note Unlike sd_watchdog_enabled(3), this method does not mutate the
+    #   environment.
+    def self.watchdog?
+      wd_usec = ENV["WATCHDOG_USEC"]
+      wd_pid = ENV["WATCHDOG_PID"]
+
+      return false unless wd_usec
+
+      begin
+        wd_usec = Integer(wd_usec)
+      rescue
+        return false
+      end
+
+      return false if wd_usec <= 0
+      return true if !wd_pid || wd_pid == $$.to_s
+
+      false
+    end
+
+    # Notify systemd with the provided state, via the notification socket, if
+    # any.
+    #
+    # Generally this method will be used indirectly through the other methods
+    # of the library.
+    #
+    # @param state [String]
+    # @param unset_env [Boolean]
+    #
+    # @return [Fixnum, nil] the number of bytes written to the notification
+    #   socket or nil if there was no socket to report to (eg. the program wasn't
+    #   started by systemd)
+    #
+    # @raise [NotifyError] if there was an error communicating with the systemd
+    #   socket
+    #
+    # @see https://www.freedesktop.org/software/systemd/man/sd_notify.html
+    def self.notify(state, unset_env = false)
+      sock = ENV["NOTIFY_SOCKET"]
+
+      return nil unless sock
+
+      ENV.delete("NOTIFY_SOCKET") if unset_env
+
+      begin
+        Addrinfo.unix(sock, :DGRAM).connect do |s|
+          s.close_on_exec = true
+          s.write(state)
+        end
+      rescue => e
+        raise NotifyError, "#{e.class}: #{e.message}", e.backtrace
+      end
+    end
+  end
+end

data/lib/sidekiq/systemd.rb

@@ -0,0 +1,24 @@
+#
+# Sidekiq's systemd integration allows Sidekiq to inform systemd:
+#  1. when it has successfully started
+#  2. when it is starting shutdown
+#  3. periodically for a liveness check with a watchdog thread
+#
+module Sidekiq
+  def self.start_watchdog
+    usec = Integer(ENV["WATCHDOG_USEC"])
+    return Sidekiq.logger.error("systemd Watchdog too fast: " + usec) if usec < 1_000_000
+
+    sec_f = usec / 1_000_000.0
+    # "It is recommended that a daemon sends a keep-alive notification message
+    # to the service manager every half of the time returned here."
+    ping_f = sec_f / 2
+    Sidekiq.logger.info "Pinging systemd watchdog every #{ping_f.round(1)} sec"
+    Thread.new do
+      loop do
+        sleep ping_f
+        Sidekiq::SdNotify.watchdog
+      end
+    end
+  end
+end

data/lib/sidekiq/testing/inline.rb

@@ -1,5 +1,6 @@
 # frozen_string_literal: true
-require 'sidekiq/testing'
+
+require "sidekiq/testing"
 
 ##
 # The Sidekiq inline infrastructure overrides perform_async so that it