sidekiq 5.0.0 → 6.0.0

data/lib/sidekiq/fetch.rb

@@ -1,5 +1,6 @@
 # frozen_string_literal: true
-require 'sidekiq'
+
+require "sidekiq"
 
 module Sidekiq
   class BasicFetch
@@ -7,13 +8,13 @@ module Sidekiq
     # can check if the process is shutting down.
     TIMEOUT = 2
 
-    UnitOfWork = Struct.new(:queue, :job) do
+    UnitOfWork = Struct.new(:queue, :job) {
       def acknowledge
         # nothing to do
       end
 
       def queue_name
-        queue.sub(/.*queue:/, ''.freeze)
+        queue.sub(/.*queue:/, "")
       end
 
       def requeue
@@ -21,7 +22,7 @@ module Sidekiq
           conn.rpush("queue:#{queue_name}", job)
         end
       end
-    end
+    }
 
     def initialize(options)
       @strictly_ordered_queues = !!options[:strict]
@@ -52,7 +53,6 @@ module Sidekiq
       end
     end
 
-
     # By leaving this as a class method, it can be pluggable and used by the Manager actor. Making it
     # an instance method will make it async to the Fetcher actor
     def self.bulk_requeue(inprogress, options)
@@ -76,6 +76,5 @@ module Sidekiq
     rescue => ex
       Sidekiq.logger.warn("Failed to requeue #{inprogress.size} jobs: #{ex.message}")
     end
-
   end
 end

data/lib/sidekiq/job_logger.rb

@@ -1,27 +1,55 @@
+# frozen_string_literal: true
+
 module Sidekiq
   class JobLogger
+    def initialize(logger = Sidekiq.logger)
+      @logger = logger
+    end
 
     def call(item, queue)
-      begin
-        start = Time.now
-        logger.info("start".freeze)
-        yield
-        logger.info("done: #{elapsed(start)} sec")
-      rescue Exception
-        logger.info("fail: #{elapsed(start)} sec")
-        raise
+      start = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
+      @logger.info("start")
+
+      yield
+
+      with_elapsed_time_context(start) do
+        @logger.info("done")
+      end
+    rescue Exception
+      with_elapsed_time_context(start) do
+        @logger.info("fail")
       end
+
+      raise
     end
 
-    private
+    def with_job_hash_context(job_hash, &block)
+      @logger.with_context(job_hash_context(job_hash), &block)
+    end
 
-    def elapsed(start)
-      (Time.now - start).round(3)
+    def job_hash_context(job_hash)
+      # If we're using a wrapper class, like ActiveJob, use the "wrapped"
+      # attribute to expose the underlying thing.
+      h = {
+        class: job_hash["wrapped"] || job_hash["class"],
+        jid: job_hash["jid"],
+      }
+      h[:bid] = job_hash["bid"] if job_hash["bid"]
+      h
     end
 
-    def logger
-      Sidekiq.logger
+    def with_elapsed_time_context(start, &block)
+      @logger.with_context(elapsed_time_context(start), &block)
+    end
+
+    def elapsed_time_context(start)
+      {elapsed: elapsed(start).to_s}
+    end
+
+    private
+
+    def elapsed(start)
+      (::Process.clock_gettime(::Process::CLOCK_MONOTONIC) - start).round(3)
     end
   end
 end
-

data/lib/sidekiq/job_retry.rb

@@ -1,5 +1,7 @@
-require 'sidekiq/scheduled'
-require 'sidekiq/api'
+# frozen_string_literal: true
+
+require "sidekiq/scheduled"
+require "sidekiq/api"
 
 module Sidekiq
   ##
@@ -55,7 +57,8 @@ module Sidekiq
   #    end
   #
   class JobRetry
-    class Skip < ::RuntimeError; end
+    class Handled < ::RuntimeError; end
+    class Skip < Handled; end
 
     include Sidekiq::Util
 
@@ -70,7 +73,7 @@ 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
@@ -79,11 +82,18 @@ 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
-    end
+      if msg["retry"]
+        attempt_retry(nil, msg, queue, e)
+      else
+        Sidekiq.death_handlers.each do |handler|
+          handler.call(msg, e)
+        rescue => handler_ex
+          handle_exception(handler_ex, {context: "Error calling death handler", job: msg})
+        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.
@@ -95,7 +105,7 @@ 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
@@ -104,11 +114,11 @@ module Sidekiq
       # 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']
+      if msg["retry"].nil?
+        msg["retry"] = worker.class.get_sidekiq_options["retry"]
       end
 
-      raise e unless msg['retry']
+      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
@@ -121,47 +131,42 @@ module Sidekiq
     # 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)
+      max_retry_attempts = retry_attempts_from(msg["retry"], @max_retries)
 
-      msg['queue'] = if msg['retry_queue']
-        msg['retry_queue']
-      else
-        queue
-      end
+      msg["queue"] = (msg["retry_queue"] || queue)
 
-      # 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!
       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
+      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
+        msg["failed_at"] = Time.now.to_f
+        msg["retry_count"] = 0
       end
 
-      if msg['backtrace'] == true
-        msg['error_backtrace'] = exception.backtrace
-      elsif !msg['backtrace']
+      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]
+      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)
-        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|
-          conn.zadd('retry', retry_at.to_s, payload)
+          conn.zadd("retry", retry_at.to_s, payload)
         end
       else
         # Goodbye dear message, you (re)tried your best I'm sure.
@@ -170,28 +175,26 @@ module Sidekiq
     end
 
     def retries_exhausted(worker, msg, exception)
-      logger.debug { "Retries exhausted for job" }
       begin
-        block = worker && worker.sidekiq_retries_exhausted_block || Sidekiq.default_retries_exhausted
-        block.call(msg, exception) if block
+        block = worker&.sidekiq_retries_exhausted_block
+        block&.call(msg, exception)
+      rescue => e
+        handle_exception(e, {context: "Error calling retries_exhausted", job: msg})
+      end
+
+      Sidekiq.death_handlers.each do |handler|
+        handler.call(msg, exception)
       rescue => e
-        handle_exception(e, { context: "Error calling retries_exhausted for #{msg['class']}", job: msg })
+        handle_exception(e, {context: "Error calling death handler", job: msg})
       end
 
-      send_to_morgue(msg) unless msg['dead'] == false
+      send_to_morgue(msg) unless msg["dead"] == false
     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)
-      now = Time.now.to_f
-      Sidekiq.redis do |conn|
-        conn.multi do
-          conn.zadd('dead', now, payload)
-          conn.zremrangebyscore('dead', '-inf', now - DeadSet.timeout)
-          conn.zremrangebyrank('dead', 0, -DeadSet.max_jobs)
-        end
-      end
+      DeadSet.new.kill(payload, notify_failure: false)
     end
 
     def retry_attempts_from(msg_retry, default)
@@ -203,21 +206,23 @@ 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&.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))
+      (count**4) + 15 + (rand(30) * (count + 1))
     end
 
     def retry_in(worker, count, exception)
-      begin
-        worker.sidekiq_retry_in_block.call(count, exception).to_i
-      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
+      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
 
     def exception_caused_by_shutdown?(e, checked_causes = [])
@@ -231,5 +236,14 @@ 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)
+      # 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 !!!"
+    end
   end
 end

data/lib/sidekiq/launcher.rb

@@ -1,17 +1,24 @@
-# encoding: utf-8
 # frozen_string_literal: true
-require 'sidekiq/manager'
-require 'sidekiq/fetch'
-require 'sidekiq/scheduled'
+
+require "sidekiq/manager"
+require "sidekiq/fetch"
+require "sidekiq/scheduled"
 
 module Sidekiq
-  # The Launcher is a very simple Actor whose job is to
-  # start, monitor and stop the core Actors in Sidekiq.
-  # If any of these actors die, the Sidekiq process exits
-  # immediately.
+  # The Launcher starts the Manager and Poller threads and provides the process heartbeat.
   class Launcher
     include Util
 
+    STATS_TTL = 5 * 365 * 24 * 60 * 60 # 5 years
+
+    PROCTITLES = [
+      proc { "sidekiq" },
+      proc { Sidekiq::VERSION },
+      proc { |me, data| data["tag"] },
+      proc { |me, data| "[#{Processor::WORKER_STATE.size} of #{data["concurrency"]} busy]" },
+      proc { |me, data| "stopping" if me.stopping? },
+    ]
+
     attr_accessor :manager, :poller, :fetcher
 
     def initialize(options)
@@ -39,7 +46,7 @@ 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
@@ -61,10 +68,30 @@ module Sidekiq
 
     private unless $TESTING
 
+    def start_heartbeat
+      loop do
+        heartbeat
+        sleep 5
+      end
+      Sidekiq.logger.info("Heartbeat stopping...")
+    end
+
+    def clear_heartbeat
+      # Remove record from Redis since we are shutting down.
+      # Note we don't stop the heartbeat thread; if the process
+      # doesn't actually exit, it'll reappear in the Web UI.
+      Sidekiq.redis do |conn|
+        conn.pipelined do
+          conn.srem("processes", identity)
+          conn.del("#{identity}:workers")
+        end
+      end
+    rescue
+      # best effort, ignore network errors
+    end
+
     def heartbeat
-      results = Sidekiq::CLI::PROCTITLES.map {|x| x.(self, to_data) }
-      results.compact!
-      $0 = results.join(' ')
+      $0 = PROCTITLES.map { |proc| proc.call(self, to_data) }.compact.join(" ")
 
       ❤
     end
@@ -72,71 +99,73 @@ module Sidekiq
     def ❤
       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"
+        nowdate = Time.now.utc.strftime("%Y-%m-%d")
 
-        workers_key = "#{key}:workers".freeze
-        nowdate = Time.now.utc.strftime("%Y-%m-%d".freeze)
         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)
           end
         end
+
         fails = procd = 0
 
-        _, exists, _, _, msg = Sidekiq.redis do |conn|
-          conn.multi do
-            conn.sadd('processes', key)
+        _, exists, _, _, msg = Sidekiq.redis { |conn|
+          res = conn.multi {
+            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
-        end
+          }
+
+          res
+        }
 
         # first heartbeat or recovering from an outage and need to reestablish our heartbeat
-        fire_event(:heartbeat) if !exists
+        fire_event(:heartbeat) unless exists
 
         return unless msg
 
-        ::Process.kill(msg, $$)
+        ::Process.kill(msg, ::Process.pid)
       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)
-      end
-    end
-
-    def start_heartbeat
-      while true
-        heartbeat
-        sleep 5
+        Processor::PROCESSED.incr(procd)
+        Processor::FAILURE.incr(fails)
       end
-      Sidekiq.logger.info("Heartbeat stopping...")
     end
 
     def to_data
       @data ||= begin
         {
-          'hostname' => hostname,
-          'started_at' => Time.now.to_f,
-          'pid' => $$,
-          'tag' => @options[:tag] || '',
-          'concurrency' => @options[:concurrency],
-          'queues' => @options[:queues].uniq,
-          'labels' => @options[:labels],
-          'identity' => identity,
+          "hostname" => hostname,
+          "started_at" => Time.now.to_f,
+          "pid" => ::Process.pid,
+          "tag" => @options[:tag] || "",
+          "concurrency" => @options[:concurrency],
+          "queues" => @options[:queues].uniq,
+          "labels" => @options[:labels],
+          "identity" => identity,
         }
       end
     end
@@ -148,20 +177,5 @@ module Sidekiq
         Sidekiq.dump_json(to_data)
       end
     end
-
-    def clear_heartbeat
-      # Remove record from Redis since we are shutting down.
-      # Note we don't stop the heartbeat thread; if the process
-      # doesn't actually exit, it'll reappear in the Web UI.
-      Sidekiq.redis do |conn|
-        conn.pipelined do
-          conn.srem('processes', identity)
-          conn.del("#{identity}:workers")
-        end
-      end
-    rescue
-      # best effort, ignore network errors
-    end
-
   end
 end