sidekiq 5.2.1 → 6.4.0

data/lib/sidekiq/job_logger.rb

@@ -1,25 +1,63 @@
 # frozen_string_literal: true
+
 module Sidekiq
   class JobLogger
+    def initialize(logger = Sidekiq.logger)
+      @logger = logger
+    end
 
     def call(item, queue)
-      start = Time.now
-      logger.info("start")
+      start = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
+      @logger.info("start")
+
       yield
-      logger.info("done: #{elapsed(start)} sec")
+
+      with_elapsed_time_context(start) do
+        @logger.info("done")
+      end
     rescue Exception
-      logger.info("fail: #{elapsed(start)} sec")
+      with_elapsed_time_context(start) do
+        @logger.info("fail")
+      end
+
       raise
     end
 
-    private
+    def prepare(job_hash, &block)
+      level = job_hash["log_level"]
+      if level
+        @logger.log_at(level) do
+          Sidekiq::Context.with(job_hash_context(job_hash), &block)
+        end
+      else
+        Sidekiq::Context.with(job_hash_context(job_hash), &block)
+      end
+    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["display_class"] || job_hash["wrapped"] || job_hash["class"],
+        jid: job_hash["jid"]
+      }
+      h[:bid] = job_hash["bid"] if job_hash["bid"]
+      h[:tags] = job_hash["tags"] if job_hash["tags"]
+      h
     end
 
-    def logger
-      Sidekiq.logger
+    def with_elapsed_time_context(start, &block)
+      Sidekiq::Context.with(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

@@ -1,6 +1,10 @@
 # frozen_string_literal: true
-require 'sidekiq/scheduled'
-require 'sidekiq/api'
+
+require "sidekiq/scheduled"
+require "sidekiq/api"
+
+require "zlib"
+require "base64"
 
 module Sidekiq
   ##
@@ -30,9 +34,10 @@ module Sidekiq
   # 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:
+  # Relevant options for job retries:
   #
-  #  * 'queue' - the queue to use
+  #  * 'queue' - the queue for the initial job
+  #  * 'retry_queue' - if job retries should be pushed to a different (e.g. lower priority) queue
   #  * 'retry_count' - number of times we've retried so far.
   #  * 'error_message' - the message from the exception
   #  * 'error_class' - the exception class
@@ -48,15 +53,18 @@ module Sidekiq
   #
   #   Sidekiq.options[:max_retries] = 7
   #
-  # or limit the number of retries for a particular worker with:
+  # or limit the number of retries for a particular worker and send retries to
+  # a low priority queue with:
   #
   #    class MyWorker
   #      include Sidekiq::Worker
-  #      sidekiq_options :retry => 10
+  #      sidekiq_options retry: 10, retry_queue: 'low'
   #    end
   #
   class JobRetry
-    class Skip < ::RuntimeError; end
+    class Handled < ::RuntimeError; end
+
+    class Skip < Handled; end
 
     include Sidekiq::Util
 
@@ -69,9 +77,9 @@ module Sidekiq
     # 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)
+    def global(jobstr, queue)
       yield
-    rescue Skip => ex
+    rescue Handled => ex
       raise ex
     rescue Sidekiq::Shutdown => ey
       # ignore, will be pushed back onto queue during hard_shutdown
@@ -80,11 +88,19 @@ 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
-    end
+      msg = Sidekiq.load_json(jobstr)
+      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.
@@ -94,9 +110,9 @@ module Sidekiq
     # 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)
+    def local(worker, jobstr, queue)
       yield
-    rescue Skip => ex
+    rescue Handled => ex
       raise ex
     rescue Sidekiq::Shutdown => ey
       # ignore, will be pushed back onto queue during hard_shutdown
@@ -105,11 +121,12 @@ 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']
+      msg = Sidekiq.load_json(jobstr)
+      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
@@ -122,47 +139,44 @@ 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']
-        # do nothing
-      elsif msg['backtrace'].to_i != 0
-        msg['error_backtrace'] = exception.backtrace[0...msg['backtrace'].to_i]
+      if msg["backtrace"]
+        lines = if msg["backtrace"] == true
+          exception.backtrace
+        else
+          exception.backtrace[0...msg["backtrace"].to_i]
+        end
+
+        msg["error_backtrace"] = compress_backtrace(lines)
       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.
@@ -171,27 +185,24 @@ 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
+        block = worker&.sidekiq_retries_exhausted_block
+        block&.call(msg, exception)
       rescue => e
-        handle_exception(e, { context: "Error calling retries_exhausted", job: msg })
+        handle_exception(e, {context: "Error calling retries_exhausted", job: msg})
       end
 
+      send_to_morgue(msg) unless msg["dead"] == false
+
       Sidekiq.death_handlers.each do |handler|
-        begin
-          handler.call(msg, exception)
-        rescue => e
-          handle_exception(e, { context: "Error calling death handler", job: msg })
-        end
+        handler.call(msg, exception)
+      rescue => e
+        handle_exception(e, {context: "Error calling death handler", job: msg})
       end
-
-      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)
       DeadSet.new.kill(payload, notify_failure: false)
     end
@@ -205,25 +216,19 @@ module Sidekiq
     end
 
     def delay_for(worker, count, exception)
-      if worker && worker.sidekiq_retry_in_block
+      jitter = rand(10) * (count + 1)
+      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
+        return custom_retry_in + jitter 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 + jitter
     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
+      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 = [])
@@ -237,5 +242,20 @@ 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
+
+    def compress_backtrace(backtrace)
+      serialized = Sidekiq.dump_json(backtrace)
+      compressed = Zlib::Deflate.deflate(serialized)
+      Base64.encode64(compressed)
+    end
   end
 end

data/lib/sidekiq/job_util.rb

@@ -0,0 +1,65 @@
+require "securerandom"
+require "time"
+
+module Sidekiq
+  module JobUtil
+    # These functions encapsulate various job utilities.
+    # They must be simple and free from side effects.
+
+    def validate(item)
+      raise(ArgumentError, "Job must be a Hash with 'class' and 'args' keys: `#{item}`") unless item.is_a?(Hash) && item.key?("class") && item.key?("args")
+      raise(ArgumentError, "Job args must be an Array: `#{item}`") unless item["args"].is_a?(Array)
+      raise(ArgumentError, "Job class must be either a Class or String representation of the class name: `#{item}`") unless item["class"].is_a?(Class) || item["class"].is_a?(String)
+      raise(ArgumentError, "Job 'at' must be a Numeric timestamp: `#{item}`") if item.key?("at") && !item["at"].is_a?(Numeric)
+      raise(ArgumentError, "Job tags must be an Array: `#{item}`") if item["tags"] && !item["tags"].is_a?(Array)
+
+      if Sidekiq.options[:on_complex_arguments] == :raise
+        msg = <<~EOM
+          Job arguments to #{item["class"]} must be native JSON types, see https://github.com/mperham/sidekiq/wiki/Best-Practices.
+          To disable this error, remove `Sidekiq.strict_args!` from your initializer.
+        EOM
+        raise(ArgumentError, msg) unless json_safe?(item)
+      elsif Sidekiq.options[:on_complex_arguments] == :warn
+        Sidekiq.logger.warn <<~EOM unless json_safe?(item)
+          Job arguments to #{item["class"]} do not serialize to JSON safely. This will raise an error in
+          Sidekiq 7.0. See https://github.com/mperham/sidekiq/wiki/Best-Practices or raise an error today
+          by calling `Sidekiq.strict_args!` during Sidekiq initialization.
+        EOM
+      end
+    end
+
+    def normalize_item(item)
+      validate(item)
+
+      # merge in the default sidekiq_options for the item's class and/or wrapped element
+      # this allows ActiveJobs to control sidekiq_options too.
+      defaults = normalized_hash(item["class"])
+      defaults = defaults.merge(item["wrapped"].get_sidekiq_options) if item["wrapped"].respond_to?(:get_sidekiq_options)
+      item = defaults.merge(item)
+
+      raise(ArgumentError, "Job must include a valid queue name") if item["queue"].nil? || item["queue"] == ""
+
+      item["class"] = item["class"].to_s
+      item["queue"] = item["queue"].to_s
+      item["jid"] ||= SecureRandom.hex(12)
+      item["created_at"] ||= Time.now.to_f
+
+      item
+    end
+
+    def normalized_hash(item_class)
+      if item_class.is_a?(Class)
+        raise(ArgumentError, "Message must include a Sidekiq::Worker class, not class name: #{item_class.ancestors.inspect}") unless item_class.respond_to?(:get_sidekiq_options)
+        item_class.get_sidekiq_options
+      else
+        Sidekiq.default_worker_options
+      end
+    end
+
+    private
+
+    def json_safe?(item)
+      JSON.parse(JSON.dump(item["args"])) == item["args"]
+    end
+  end
+end