sidekiq 5.1.1 → 5.2.9

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

@@ -1,3 +1,4 @@
+# frozen_string_literal: true
 module Sidekiq
   module Extensions
 

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

@@ -1,9 +1,10 @@
+# frozen_string_literal: true
 module Sidekiq
   class JobLogger
 
     def call(item, queue)
-      start = Time.now
-      logger.info("start".freeze)
+      start = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
+      logger.info("start")
       yield
       logger.info("done: #{elapsed(start)} sec")
     rescue Exception
@@ -14,7 +15,7 @@ module Sidekiq
     private
 
     def elapsed(start)
-      (Time.now - start).round(3)
+      (::Process.clock_gettime(::Process::CLOCK_MONOTONIC) - start).round(3)
     end
 
     def logger

data/lib/sidekiq/job_retry.rb

@@ -1,3 +1,4 @@
+# frozen_string_literal: true
 require 'sidekiq/scheduled'
 require 'sidekiq/api'
 
@@ -55,7 +56,8 @@ module Sidekiq
   #    end
   #
   class JobRetry
-    class Skip < ::RuntimeError; end
+    class Handled < ::RuntimeError; end
+    class Skip < Handled; end
 
     include Sidekiq::Util
 
@@ -70,7 +72,7 @@ module Sidekiq
     # require the worker to be instantiated.
     def global(msg, queue)
       yield
-    rescue Skip => ex
+    rescue Handled => ex
       raise ex
     rescue Sidekiq::Shutdown => ey
       # ignore, will be pushed back onto queue during hard_shutdown
@@ -79,9 +81,19 @@ module Sidekiq
       # ignore, will be pushed back onto queue during hard_shutdown
       raise Sidekiq::Shutdown if exception_caused_by_shutdown?(e)
 
-      raise e unless msg['retry']
-      attempt_retry(nil, msg, queue, e)
-      raise 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
 
 
@@ -95,7 +107,7 @@ module Sidekiq
     # calling the handle_exception handlers.
     def local(worker, msg, queue)
       yield
-    rescue Skip => ex
+    rescue Handled => ex
       raise ex
     rescue Sidekiq::Shutdown => ey
       # ignore, will be pushed back onto queue during hard_shutdown
@@ -129,9 +141,7 @@ module Sidekiq
         queue
       end
 
-      # App code can stuff all sorts of crazy binary data into the error message
-      # that won't convert to JSON.
-      m = exception.message.to_s[0, 10_000]
+      m = exception_message(exception)
       if m.respond_to?(:scrub!)
         m.force_encoding("utf-8")
         m.scrub!
@@ -157,7 +167,8 @@ module Sidekiq
 
       if count < max_retry_attempts
         delay = delay_for(worker, count, exception)
-        logger.debug { "Failure! Retry #{count} in #{delay} seconds" }
+        # 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|
@@ -170,7 +181,6 @@ module Sidekiq
     end
 
     def retries_exhausted(worker, msg, exception)
-      logger.debug { "Retries exhausted for job" }
       begin
         block = worker && worker.sidekiq_retries_exhausted_block
         block.call(msg, exception) if block
@@ -190,7 +200,7 @@ module Sidekiq
     end
 
     def send_to_morgue(msg)
-      Sidekiq.logger.info { "Adding dead #{msg['class']} job #{msg['jid']}" }
+      logger.info { "Adding dead #{msg['class']} job #{msg['jid']}" }
       payload = Sidekiq.dump_json(msg)
       DeadSet.new.kill(payload, notify_failure: false)
     end
@@ -204,7 +214,11 @@ module Sidekiq
     end
 
     def delay_for(worker, count, exception)
-      worker && worker.sidekiq_retry_in_block && retry_in(worker, count, exception) || seconds_to_delay(count)
+      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
@@ -214,7 +228,7 @@ module Sidekiq
 
     def retry_in(worker, count, exception)
       begin
-        worker.sidekiq_retry_in_block.call(count, exception).to_i
+        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
@@ -232,5 +246,17 @@ module Sidekiq
         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
@@ -73,19 +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)
@@ -97,7 +103,7 @@ module Sidekiq
           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.hmset(key, 'info', to_json, 'busy', curstate.size, 'beat', Time.now.to_f, 'quiet', @done)
             conn.expire(key, 60)
             conn.rpop("#{key}-signals")
           end
@@ -113,8 +119,8 @@ module Sidekiq
         # 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
 

data/lib/sidekiq/logging.rb

@@ -33,9 +33,9 @@ module Sidekiq
     def self.job_hash_context(job_hash)
       # If we're using a wrapper class, like ActiveJob, use the "wrapped"
       # attribute to expose the underlying thing.
-      klass = job_hash['wrapped'.freeze] || job_hash["class".freeze]
-      bid = job_hash['bid'.freeze]
-      "#{klass} JID-#{job_hash['jid'.freeze]}#{" BID-#{bid}" if bid}"
+      klass = job_hash['wrapped'] || job_hash["class"]
+      bid = job_hash['bid']
+      "#{klass} JID-#{job_hash['jid']}#{" BID-#{bid}" if bid}"
     end
 
     def self.with_job_hash_context(job_hash, &block)

data/lib/sidekiq/manager.rb

@@ -1,4 +1,3 @@
-# encoding: utf-8
 # frozen_string_literal: true
 require 'sidekiq/util'
 require 'sidekiq/processor'
@@ -31,7 +30,7 @@ module Sidekiq
     def initialize(options={})
       logger.debug { options.inspect }
       @options = options
-      @count = options[:concurrency] || 25
+      @count = options[:concurrency] || 10
       raise ArgumentError, "Concurrency of #{@count} is not supported" if @count < 1
 
       @done = false
@@ -71,11 +70,11 @@ module Sidekiq
       return if @workers.empty?
 
       logger.info { "Pausing to allow workers to finish..." }
-      remaining = deadline - Time.now
+      remaining = deadline - ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
       while remaining > PAUSE_TIME
         return if @workers.empty?
         sleep PAUSE_TIME
-        remaining = deadline - Time.now
+        remaining = deadline - ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
       end
       return if @workers.empty?
 

data/lib/sidekiq/middleware/server/active_record.rb

@@ -1,3 +1,4 @@
+# frozen_string_literal: true
 module Sidekiq
   module Middleware
     module Server
@@ -6,7 +7,7 @@ module Sidekiq
         def initialize
           # With Rails 5+ we must use the Reloader **always**.
           # The reloader handles code loading and db connection management.
-          if ::Rails::VERSION::MAJOR >= 5
+          if defined?(::Rails) && defined?(::Rails::VERSION) && ::Rails::VERSION::MAJOR >= 5
             raise ArgumentError, "Rails 5 no longer needs or uses the ActiveRecord middleware."
           end
         end

data/lib/sidekiq/processor.rb

@@ -4,8 +4,6 @@ require 'sidekiq/fetch'
 require 'sidekiq/job_logger'
 require 'sidekiq/job_retry'
 require 'thread'
-require 'concurrent/map'
-require 'concurrent/atomic/atomic_fixnum'
 
 module Sidekiq
   ##
@@ -89,7 +87,7 @@ module Sidekiq
     def get_one
       begin
         work = @strategy.retrieve_work
-        (logger.info { "Redis is online, #{Time.now - @down} sec downtime" }; @down = nil) if @down
+        (logger.info { "Redis is online, #{::Process.clock_gettime(::Process::CLOCK_MONOTONIC) - @down} sec downtime" }; @down = nil) if @down
         work
       rescue Sidekiq::Shutdown
       rescue => ex
@@ -109,7 +107,7 @@ module Sidekiq
 
     def handle_fetch_exception(ex)
       if !@down
-        @down = Time.now
+        @down = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
         logger.error("Error fetching job: #{ex}")
         handle_exception(ex)
       end
@@ -132,9 +130,9 @@ module Sidekiq
               # the Reloader.  It handles code loading, db connection management, etc.
               # Effectively this block denotes a "unit of work" to Rails.
               @reloader.call do
-                klass  = constantize(job_hash['class'.freeze])
+                klass  = constantize(job_hash['class'])
                 worker = klass.new
-                worker.jid = job_hash['jid'.freeze]
+                worker.jid = job_hash['jid']
                 @retrier.local(worker, pristine, queue) do
                   yield worker
                 end
@@ -149,24 +147,22 @@ module Sidekiq
       jobstr = work.job
       queue = work.queue_name
 
-      ack = false
+      # Treat malformed JSON as a special case: job goes straight to the morgue.
+      job_hash = nil
       begin
-        # Treat malformed JSON as a special case: job goes straight to the morgue.
-        job_hash = nil
-        begin
-          job_hash = Sidekiq.load_json(jobstr)
-        rescue => ex
-          handle_exception(ex, { :context => "Invalid JSON for job", :jobstr => jobstr })
-          # we can't notify because the job isn't a valid hash payload.
-          DeadSet.new.kill(jobstr, notify_failure: false)
-          ack = true
-          raise
-        end
+        job_hash = Sidekiq.load_json(jobstr)
+      rescue => ex
+        handle_exception(ex, { :context => "Invalid JSON for job", :jobstr => jobstr })
+        # we can't notify because the job isn't a valid hash payload.
+        DeadSet.new.kill(jobstr, notify_failure: false)
+        return work.acknowledge
+      end
 
-        ack = true
+      ack = true
+      begin
         dispatch(job_hash, queue) do |worker|
           Sidekiq.server_middleware.invoke(worker, job_hash, queue) do
-            execute_job(worker, cloned(job_hash['args'.freeze]))
+            execute_job(worker, cloned(job_hash['args']))
           end
         end
       rescue Sidekiq::Shutdown
@@ -174,10 +170,19 @@ module Sidekiq
         # within the timeout.  Don't acknowledge the work since
         # we didn't properly finish it.
         ack = false
-      rescue Exception => ex
-        e = ex.is_a?(::Sidekiq::JobRetry::Skip) && ex.cause ? ex.cause : ex
+      rescue Sidekiq::JobRetry::Handled => h
+        # this is the common case: job raised error and Sidekiq::JobRetry::Handled
+        # signals that we created a retry successfully.  We can acknowlege the job.
+        e = h.cause ? h.cause : h
         handle_exception(e, { :context => "Job raised exception", :job => job_hash, :jobstr => jobstr })
         raise e
+      rescue Exception => ex
+        # Unexpected error!  This is very bad and indicates an exception that got past
+        # the retry subsystem (e.g. network partition).  We won't acknowledge the job
+        # so it can be rescued when using Sidekiq Pro.
+        ack = false
+        handle_exception(ex, { :context => "Internal exception!", :job => job_hash, :jobstr => jobstr })
+        raise e
       ensure
         work.acknowledge if ack
       end
@@ -187,22 +192,68 @@ module Sidekiq
       worker.perform(*cloned_args)
     end
 
-    WORKER_STATE = Concurrent::Map.new
-    PROCESSED = Concurrent::AtomicFixnum.new
-    FAILURE = Concurrent::AtomicFixnum.new
+    # Ruby doesn't provide atomic counters out of the box so we'll
+    # implement something simple ourselves.
+    # https://bugs.ruby-lang.org/issues/14706
+    class Counter
+      def initialize
+        @value = 0
+        @lock = Mutex.new
+      end
+
+      def incr(amount=1)
+        @lock.synchronize { @value = @value + amount }
+      end
+
+      def reset
+        @lock.synchronize { val = @value; @value = 0; val }
+      end
+    end
+
+    # jruby's Hash implementation is not threadsafe, so we wrap it in a mutex here
+    class SharedWorkerState
+      def initialize
+        @worker_state = {}
+        @lock = Mutex.new
+      end
+
+      def set(tid, hash)
+        @lock.synchronize { @worker_state[tid] = hash }
+      end
+
+      def delete(tid)
+        @lock.synchronize { @worker_state.delete(tid) }
+      end
+
+      def dup
+        @lock.synchronize { @worker_state.dup }
+      end
+
+      def size
+        @lock.synchronize { @worker_state.size }
+      end
+
+      def clear
+        @lock.synchronize { @worker_state.clear }
+      end
+    end
+
+    PROCESSED = Counter.new
+    FAILURE = Counter.new
+    WORKER_STATE = SharedWorkerState.new
 
     def stats(job_hash, queue)
       tid = Sidekiq::Logging.tid
-      WORKER_STATE[tid] = {:queue => queue, :payload => job_hash, :run_at => Time.now.to_i }
+      WORKER_STATE.set(tid, {:queue => queue, :payload => job_hash, :run_at => Time.now.to_i })
 
       begin
         yield
       rescue Exception
-        FAILURE.increment
+        FAILURE.incr
         raise
       ensure
         WORKER_STATE.delete(tid)
-        PROCESSED.increment
+        PROCESSED.incr
       end
     end