sidekiq 4.2.10 → 5.2.10

data/lib/sidekiq/ctl.rb

@@ -0,0 +1,221 @@
+#!/usr/bin/env ruby
+
+require 'fileutils'
+require 'sidekiq/api'
+
+class Sidekiq::Ctl
+  DEFAULT_KILL_TIMEOUT = 10
+  CMD = File.basename($0)
+
+  attr_reader :stage, :pidfile, :kill_timeout
+
+  def self.print_usage
+    puts "#{CMD} - control Sidekiq from the command line."
+    puts
+    puts "Usage: #{CMD} quiet <pidfile> <kill_timeout>"
+    puts "       #{CMD} stop <pidfile> <kill_timeout>"
+    puts "       #{CMD} status <section>"
+    puts
+    puts "       <pidfile> is path to a pidfile"
+    puts "       <kill_timeout> is number of seconds to wait until Sidekiq exits"
+    puts "       (default: #{Sidekiq::Ctl::DEFAULT_KILL_TIMEOUT}), after which Sidekiq will be KILL'd"
+    puts
+    puts "       <section> (optional) view a specific section of the status output"
+    puts "       Valid sections are: #{Sidekiq::Ctl::Status::VALID_SECTIONS.join(', ')}"
+    puts
+    puts "Be sure to set the kill_timeout LONGER than Sidekiq's -t timeout.  If you want"
+    puts "to wait 60 seconds for jobs to finish, use `sidekiq -t 60` and `sidekiqctl stop"
+    puts " path_to_pidfile 61`"
+    puts
+  end
+
+  def initialize(stage, pidfile, timeout)
+    @stage = stage
+    @pidfile = pidfile
+    @kill_timeout = timeout
+
+    done('No pidfile given', :error) if !pidfile
+    done("Pidfile #{pidfile} does not exist", :warn) if !File.exist?(pidfile)
+    done('Invalid pidfile content', :error) if pid == 0
+
+    fetch_process
+
+    begin
+      send(stage)
+    rescue NoMethodError
+      done "Invalid command: #{stage}", :error
+    end
+  end
+
+  def fetch_process
+    Process.kill(0, pid)
+  rescue Errno::ESRCH
+    done "Process doesn't exist", :error
+  # We were not allowed to send a signal, but the process must have existed
+  # when Process.kill() was called.
+  rescue Errno::EPERM
+    return pid
+  end
+
+  def done(msg, error = nil)
+    puts msg
+    exit(exit_signal(error))
+  end
+
+  def exit_signal(error)
+    (error == :error) ? 1 : 0
+  end
+
+  def pid
+    @pid ||= File.read(pidfile).to_i
+  end
+
+  def quiet
+    `kill -TSTP #{pid}`
+  end
+
+  def stop
+    `kill -TERM #{pid}`
+    kill_timeout.times do
+      begin
+        Process.kill(0, pid)
+      rescue Errno::ESRCH
+        FileUtils.rm_f pidfile
+        done 'Sidekiq shut down gracefully.'
+      rescue Errno::EPERM
+        done 'Not permitted to shut down Sidekiq.'
+      end
+      sleep 1
+    end
+    `kill -9 #{pid}`
+    FileUtils.rm_f pidfile
+    done 'Sidekiq shut down forcefully.'
+  end
+  alias_method :shutdown, :stop
+
+  class Status
+    VALID_SECTIONS = %w[all version overview processes queues]
+    def display(section = nil)
+      section ||= 'all'
+      unless VALID_SECTIONS.include? section
+        puts "I don't know how to check the status of '#{section}'!"
+        puts "Try one of these: #{VALID_SECTIONS.join(', ')}"
+        return
+      end
+      send(section)
+    rescue StandardError => e
+      puts "Couldn't get status: #{e}"
+    end
+
+    def all
+      version
+      puts
+      overview
+      puts
+      processes
+      puts
+      queues
+    end
+
+    def version
+      puts "Sidekiq #{Sidekiq::VERSION}"
+      puts Time.now
+    end
+
+    def overview
+      puts '---- Overview ----'
+      puts "  Processed: #{delimit stats.processed}"
+      puts "     Failed: #{delimit stats.failed}"
+      puts "       Busy: #{delimit stats.workers_size}"
+      puts "   Enqueued: #{delimit stats.enqueued}"
+      puts "    Retries: #{delimit stats.retry_size}"
+      puts "  Scheduled: #{delimit stats.scheduled_size}"
+      puts "       Dead: #{delimit stats.dead_size}"
+    end
+
+    def processes
+      puts "---- Processes (#{process_set.size}) ----"
+      process_set.each_with_index do |process, index|
+        puts "#{process['identity']} #{tags_for(process)}"
+        puts "  Started: #{Time.at(process['started_at'])} (#{time_ago(process['started_at'])})"
+        puts "  Threads: #{process['concurrency']} (#{process['busy']} busy)"
+        puts "   Queues: #{split_multiline(process['queues'].sort, pad: 11)}"
+        puts '' unless (index+1) == process_set.size
+      end
+    end
+
+    COL_PAD = 2
+    def queues
+      puts "---- Queues (#{queue_data.size}) ----"
+      columns = {
+        name: [:ljust, (['name'] + queue_data.map(&:name)).map(&:length).max + COL_PAD],
+        size: [:rjust, (['size'] + queue_data.map(&:size)).map(&:length).max + COL_PAD],
+        latency: [:rjust, (['latency'] + queue_data.map(&:latency)).map(&:length).max + COL_PAD]
+      }
+      columns.each { |col, (dir, width)| print col.to_s.upcase.public_send(dir, width) }
+      puts
+      queue_data.each do |q|
+        columns.each do |col, (dir, width)|
+          print q.send(col).public_send(dir, width)
+        end
+        puts
+      end
+    end
+
+    private
+
+    def delimit(number)
+      number.to_s.reverse.scan(/.{1,3}/).join(',').reverse
+    end
+
+    def split_multiline(values, opts = {})
+      return 'none' unless values
+      pad = opts[:pad] || 0
+      max_length = opts[:max_length] || (80 - pad)
+      out = []
+      line = ''
+      values.each do |value|
+        if (line.length + value.length) > max_length
+          out << line
+          line = ' ' * pad
+        end
+        line << value + ', '
+      end
+      out << line[0..-3]
+      out.join("\n")
+    end
+
+    def tags_for(process)
+      tags = [
+        process['tag'],
+        process['labels'],
+        (process['quiet'] == 'true' ? 'quiet' : nil)
+      ].flatten.compact
+      tags.any? ? "[#{tags.join('] [')}]" : nil
+    end
+
+    def time_ago(timestamp)
+      seconds = Time.now - Time.at(timestamp)
+      return 'just now' if seconds < 60
+      return 'a minute ago' if seconds < 120
+      return "#{seconds.floor / 60} minutes ago" if seconds < 3600
+      return 'an hour ago' if seconds < 7200
+      "#{seconds.floor / 60 / 60} hours ago"
+    end
+
+    QUEUE_STRUCT = Struct.new(:name, :size, :latency)
+    def queue_data
+      @queue_data ||= Sidekiq::Queue.all.map do |q|
+        QUEUE_STRUCT.new(q.name, q.size.to_s, sprintf('%#.2f', q.latency))
+      end
+    end
+
+    def process_set
+      @process_set ||= Sidekiq::ProcessSet.new
+    end
+
+    def stats
+      @stats ||= Sidekiq::Stats.new
+    end
+  end
+end

data/lib/sidekiq/delay.rb

@@ -0,0 +1,42 @@
+# frozen_string_literal: true
+module Sidekiq
+  module Extensions
+
+    def self.enable_delay!
+      if defined?(::ActiveSupport)
+        require 'sidekiq/extensions/active_record'
+        require 'sidekiq/extensions/action_mailer'
+
+        # Need to patch Psych so it can autoload classes whose names are serialized
+        # in the delayed YAML.
+        Psych::Visitors::ToRuby.prepend(Sidekiq::Extensions::PsychAutoload)
+
+        ActiveSupport.on_load(:active_record) do
+          include Sidekiq::Extensions::ActiveRecord
+        end
+        ActiveSupport.on_load(:action_mailer) do
+          extend Sidekiq::Extensions::ActionMailer
+        end
+      end
+
+      require 'sidekiq/extensions/class_methods'
+      Module.__send__(:include, Sidekiq::Extensions::Klass)
+    end
+
+    module PsychAutoload
+      def resolve_class(klass_name)
+        return nil if !klass_name || klass_name.empty?
+        # constantize
+        names = klass_name.split('::')
+        names.shift if names.empty? || names.first.empty?
+
+        names.inject(Object) do |constant, name|
+          constant.const_defined?(name) ? constant.const_get(name) : constant.const_missing(name)
+        end
+      rescue NameError
+        super
+      end
+    end
+  end
+end
+

data/lib/sidekiq/exception_handler.rb

@@ -7,11 +7,10 @@ module Sidekiq
     class Logger
       def call(ex, ctxHash)
         Sidekiq.logger.warn(Sidekiq.dump_json(ctxHash)) if !ctxHash.empty?
-        Sidekiq.logger.warn "#{ex.class.name}: #{ex.message}"
-        Sidekiq.logger.warn ex.backtrace.join("\n") unless ex.backtrace.nil?
+        Sidekiq.logger.warn("#{ex.class.name}: #{ex.message}")
+        Sidekiq.logger.warn(ex.backtrace.join("\n")) unless ex.backtrace.nil?
       end
 
-      # Set up default handler which just logs the error
       Sidekiq.error_handlers << Sidekiq::ExceptionHandler::Logger.new
     end
 
@@ -26,6 +25,5 @@ module Sidekiq
         end
       end
     end
-
   end
 end

data/lib/sidekiq/extensions/generic_proxy.rb

@@ -3,6 +3,8 @@ require 'yaml'
 
 module Sidekiq
   module Extensions
+    SIZE_LIMIT = 8_192
+
     class Proxy < BasicObject
       def initialize(performable, target, options={})
         @performable = performable
@@ -17,7 +19,11 @@ module Sidekiq
         # to JSON and then deserialized on the other side back into a
         # Ruby object.
         obj = [@target, name, args]
-        @performable.client_push({ 'class' => @performable, 'args' => [::YAML.dump(obj)] }.merge(@opts))
+        marshalled = ::YAML.dump(obj)
+        if marshalled.size > SIZE_LIMIT
+          ::Sidekiq.logger.warn { "#{@target}.#{name} job argument is #{marshalled.bytesize} bytes, you should refactor it to reduce the size" }
+        end
+        @performable.client_push({ 'class' => @performable, 'args' => [marshalled] }.merge(@opts))
       end
     end
 

data/lib/sidekiq/fetch.rb

@@ -13,7 +13,7 @@ module Sidekiq
       end
 
       def queue_name
-        queue.sub(/.*queue:/, ''.freeze)
+        queue.sub(/.*queue:/, '')
       end
 
       def requeue

data/lib/sidekiq/job_logger.rb

@@ -0,0 +1,25 @@
+# frozen_string_literal: true
+module Sidekiq
+  class JobLogger
+
+    def call(item, queue)
+      start = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
+      logger.info("start")
+      yield
+      logger.info("done: #{elapsed(start)} sec")
+    rescue Exception
+      logger.info("fail: #{elapsed(start)} sec")
+      raise
+    end
+
+    private
+
+    def elapsed(start)
+      (::Process.clock_gettime(::Process::CLOCK_MONOTONIC) - start).round(3)
+    end
+
+    def logger
+      Sidekiq.logger
+    end
+  end
+end

data/lib/sidekiq/job_retry.rb

@@ -0,0 +1,262 @@
+# frozen_string_literal: true
+require 'sidekiq/scheduled'
+require 'sidekiq/api'
+
+module Sidekiq
+  ##
+  # Automatically retry jobs that fail in Sidekiq.
+  # Sidekiq's retry support assumes a typical development lifecycle:
+  #
+  #   0. Push some code changes with a bug in it.
+  #   1. Bug causes job processing to fail, Sidekiq's middleware captures
+  #      the job and pushes it onto a retry queue.
+  #   2. Sidekiq retries jobs in the retry queue multiple times with
+  #      an exponential delay, the job continues to fail.
+  #   3. After a few days, a developer deploys a fix. The job is
+  #      reprocessed successfully.
+  #   4. Once retries are exhausted, Sidekiq will give up and move the
+  #      job to the Dead Job Queue (aka morgue) where it must be dealt with
+  #      manually in the Web UI.
+  #   5. After 6 months on the DJQ, Sidekiq will discard the job.
+  #
+  # A job looks like:
+  #
+  #     { 'class' => 'HardWorker', 'args' => [1, 2, 'foo'], 'retry' => true }
+  #
+  # The 'retry' option also accepts a number (in place of 'true'):
+  #
+  #     { 'class' => 'HardWorker', 'args' => [1, 2, 'foo'], 'retry' => 5 }
+  #
+  # The job will be retried this number of times before giving up. (If simply
+  # 'true', Sidekiq retries 25 times)
+  #
+  # We'll add a bit more data to the job to support retries:
+  #
+  #  * 'queue' - the queue to use
+  #  * 'retry_count' - number of times we've retried so far.
+  #  * 'error_message' - the message from the exception
+  #  * 'error_class' - the exception class
+  #  * 'failed_at' - the first time it failed
+  #  * 'retried_at' - the last time it was retried
+  #  * 'backtrace' - the number of lines of error backtrace to store
+  #
+  # We don't store the backtrace by default as that can add a lot of overhead
+  # to the job and everyone is using an error service, right?
+  #
+  # The default number of retries is 25 which works out to about 3 weeks
+  # You can change the default maximum number of retries in your initializer:
+  #
+  #   Sidekiq.options[:max_retries] = 7
+  #
+  # or limit the number of retries for a particular worker with:
+  #
+  #    class MyWorker
+  #      include Sidekiq::Worker
+  #      sidekiq_options :retry => 10
+  #    end
+  #
+  class JobRetry
+    class Handled < ::RuntimeError; end
+    class Skip < Handled; end
+
+    include Sidekiq::Util
+
+    DEFAULT_MAX_RETRY_ATTEMPTS = 25
+
+    def initialize(options = {})
+      @max_retries = Sidekiq.options.merge(options).fetch(:max_retries, DEFAULT_MAX_RETRY_ATTEMPTS)
+    end
+
+    # The global retry handler requires only the barest of data.
+    # We want to be able to retry as much as possible so we don't
+    # require the worker to be instantiated.
+    def global(msg, queue)
+      yield
+    rescue Handled => ex
+      raise ex
+    rescue Sidekiq::Shutdown => ey
+      # ignore, will be pushed back onto queue during hard_shutdown
+      raise ey
+    rescue Exception => e
+      # ignore, will be pushed back onto queue during hard_shutdown
+      raise Sidekiq::Shutdown if exception_caused_by_shutdown?(e)
+
+      if msg['retry']
+        attempt_retry(nil, msg, queue, e)
+      else
+        Sidekiq.death_handlers.each do |handler|
+          begin
+            handler.call(msg, e)
+          rescue => handler_ex
+            handle_exception(handler_ex, { context: "Error calling death handler", job: msg })
+          end
+        end
+      end
+
+      raise Handled
+    end
+
+
+    # The local retry support means that any errors that occur within
+    # this block can be associated with the given worker instance.
+    # This is required to support the `sidekiq_retries_exhausted` block.
+    #
+    # Note that any exception from the block is wrapped in the Skip
+    # exception so the global block does not reprocess the error.  The
+    # Skip exception is unwrapped within Sidekiq::Processor#process before
+    # calling the handle_exception handlers.
+    def local(worker, msg, queue)
+      yield
+    rescue Handled => ex
+      raise ex
+    rescue Sidekiq::Shutdown => ey
+      # ignore, will be pushed back onto queue during hard_shutdown
+      raise ey
+    rescue Exception => e
+      # ignore, will be pushed back onto queue during hard_shutdown
+      raise Sidekiq::Shutdown if exception_caused_by_shutdown?(e)
+
+      if msg['retry'] == nil
+        msg['retry'] = worker.class.get_sidekiq_options['retry']
+      end
+
+      raise e unless msg['retry']
+      attempt_retry(worker, msg, queue, e)
+      # We've handled this error associated with this job, don't
+      # need to handle it at the global level
+      raise Skip
+    end
+
+    private
+
+    # Note that +worker+ can be nil here if an error is raised before we can
+    # instantiate the worker instance.  All access must be guarded and
+    # best effort.
+    def attempt_retry(worker, msg, queue, exception)
+      max_retry_attempts = retry_attempts_from(msg['retry'], @max_retries)
+
+      msg['queue'] = if msg['retry_queue']
+        msg['retry_queue']
+      else
+        queue
+      end
+
+      m = exception_message(exception)
+      if m.respond_to?(:scrub!)
+        m.force_encoding("utf-8")
+        m.scrub!
+      end
+
+      msg['error_message'] = m
+      msg['error_class'] = exception.class.name
+      count = if msg['retry_count']
+        msg['retried_at'] = Time.now.to_f
+        msg['retry_count'] += 1
+      else
+        msg['failed_at'] = Time.now.to_f
+        msg['retry_count'] = 0
+      end
+
+      if msg['backtrace'] == true
+        msg['error_backtrace'] = exception.backtrace
+      elsif !msg['backtrace']
+        # do nothing
+      elsif msg['backtrace'].to_i != 0
+        msg['error_backtrace'] = exception.backtrace[0...msg['backtrace'].to_i]
+      end
+
+      if count < max_retry_attempts
+        delay = delay_for(worker, count, exception)
+        # Logging here can break retries if the logging device raises ENOSPC #3979
+        #logger.debug { "Failure! Retry #{count} in #{delay} seconds" }
+        retry_at = Time.now.to_f + delay
+        payload = Sidekiq.dump_json(msg)
+        Sidekiq.redis do |conn|
+          conn.zadd('retry', retry_at.to_s, payload)
+        end
+      else
+        # Goodbye dear message, you (re)tried your best I'm sure.
+        retries_exhausted(worker, msg, exception)
+      end
+    end
+
+    def retries_exhausted(worker, msg, exception)
+      begin
+        block = worker && worker.sidekiq_retries_exhausted_block
+        block.call(msg, exception) if block
+      rescue => e
+        handle_exception(e, { context: "Error calling retries_exhausted", job: msg })
+      end
+
+      Sidekiq.death_handlers.each do |handler|
+        begin
+          handler.call(msg, exception)
+        rescue => e
+          handle_exception(e, { context: "Error calling death handler", job: msg })
+        end
+      end
+
+      send_to_morgue(msg) unless msg['dead'] == false
+    end
+
+    def send_to_morgue(msg)
+      logger.info { "Adding dead #{msg['class']} job #{msg['jid']}" }
+      payload = Sidekiq.dump_json(msg)
+      DeadSet.new.kill(payload, notify_failure: false)
+    end
+
+    def retry_attempts_from(msg_retry, default)
+      if msg_retry.is_a?(Integer)
+        msg_retry
+      else
+        default
+      end
+    end
+
+    def delay_for(worker, count, exception)
+      if worker && worker.sidekiq_retry_in_block
+        custom_retry_in = retry_in(worker, count, exception).to_i
+        return custom_retry_in if custom_retry_in > 0
+      end
+      seconds_to_delay(count)
+    end
+
+    # delayed_job uses the same basic formula
+    def seconds_to_delay(count)
+      (count ** 4) + 15 + (rand(30)*(count+1))
+    end
+
+    def retry_in(worker, count, exception)
+      begin
+        worker.sidekiq_retry_in_block.call(count, exception)
+      rescue Exception => e
+        handle_exception(e, { context: "Failure scheduling retry using the defined `sidekiq_retry_in` in #{worker.class.name}, falling back to default" })
+        nil
+      end
+    end
+
+    def exception_caused_by_shutdown?(e, checked_causes = [])
+      return false unless e.cause
+
+      # Handle circular causes
+      checked_causes << e.object_id
+      return false if checked_causes.include?(e.cause.object_id)
+
+      e.cause.instance_of?(Sidekiq::Shutdown) ||
+        exception_caused_by_shutdown?(e.cause, checked_causes)
+    end
+
+    # Extract message from exception.
+    # Set a default if the message raises an error
+    def exception_message(exception)
+      begin
+        # App code can stuff all sorts of crazy binary data into the error message
+        # that won't convert to JSON.
+        exception.message.to_s[0, 10_000]
+      rescue
+        "!!! ERROR MESSAGE THREW AN ERROR !!!".dup
+      end
+    end
+
+  end
+end

data/lib/sidekiq/launcher.rb

@@ -1,4 +1,3 @@
-# encoding: utf-8
 # frozen_string_literal: true
 require 'sidekiq/manager'
 require 'sidekiq/fetch'
@@ -14,6 +13,8 @@ module Sidekiq
 
     attr_accessor :manager, :poller, :fetcher
 
+    STATS_TTL = 5*365*24*60*60
+
     def initialize(options)
       @manager = Sidekiq::Manager.new(options)
       @poller = Sidekiq::Scheduled::Poller.new
@@ -39,7 +40,7 @@ module Sidekiq
     # return until all work is complete and cleaned up.
     # It can take up to the timeout to complete.
     def stop
-      deadline = Time.now + @options[:timeout]
+      deadline = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC) + @options[:timeout]
 
       @done = true
       @manager.quiet
@@ -61,8 +62,6 @@ module Sidekiq
 
     private unless $TESTING
 
-    JVM_RESERVED_SIGNALS = ['USR1', 'USR2'] # Don't Process#kill if we get these signals via the API
-
     def heartbeat
       results = Sidekiq::CLI::PROCTITLES.map {|x| x.(self, to_data) }
       results.compact!
@@ -75,19 +74,24 @@ module Sidekiq
       key = identity
       fails = procd = 0
       begin
-        Processor::FAILURE.update {|curr| fails = curr; 0 }
-        Processor::PROCESSED.update {|curr| procd = curr; 0 }
+        fails = Processor::FAILURE.reset
+        procd = Processor::PROCESSED.reset
+        curstate = Processor::WORKER_STATE.dup
 
-        workers_key = "#{key}:workers".freeze
-        nowdate = Time.now.utc.strftime("%Y-%m-%d".freeze)
+        workers_key = "#{key}:workers"
+        nowdate = Time.now.utc.strftime("%Y-%m-%d")
         Sidekiq.redis do |conn|
           conn.multi do
-            conn.incrby("stat:processed".freeze, procd)
+            conn.incrby("stat:processed", procd)
             conn.incrby("stat:processed:#{nowdate}", procd)
-            conn.incrby("stat:failed".freeze, fails)
+            conn.expire("stat:processed:#{nowdate}", STATS_TTL)
+
+            conn.incrby("stat:failed", fails)
             conn.incrby("stat:failed:#{nowdate}", fails)
+            conn.expire("stat:failed:#{nowdate}", STATS_TTL)
+
             conn.del(workers_key)
-            Processor::WORKER_STATE.each_pair do |tid, hash|
+            curstate.each_pair do |tid, hash|
               conn.hset(workers_key, tid, Sidekiq.dump_json(hash))
             end
             conn.expire(workers_key, 60)
@@ -98,8 +102,8 @@ module Sidekiq
         _, exists, _, _, msg = Sidekiq.redis do |conn|
           conn.multi do
             conn.sadd('processes', key)
-            conn.exists(key)
-            conn.hmset(key, 'info', to_json, 'busy', Processor::WORKER_STATE.size, 'beat', Time.now.to_f, 'quiet', @done)
+            conn.exists?(key)
+            conn.hmset(key, 'info', to_json, 'busy', curstate.size, 'beat', Time.now.to_f, 'quiet', @done)
             conn.expire(key, 60)
             conn.rpop("#{key}-signals")
           end
@@ -110,17 +114,13 @@ module Sidekiq
 
         return unless msg
 
-        if JVM_RESERVED_SIGNALS.include?(msg)
-          Sidekiq::CLI.instance.handle_signal(msg)
-        else
-          ::Process.kill(msg, $$)
-        end
+        ::Process.kill(msg, $$)
       rescue => e
         # ignore all redis/network issues
         logger.error("heartbeat: #{e.message}")
         # don't lose the counts if there was a network issue
-        Processor::PROCESSED.increment(procd)
-        Processor::FAILURE.increment(fails)
+        Processor::PROCESSED.incr(procd)
+        Processor::FAILURE.incr(fails)
       end
     end