Reset Active Job's execution and exception_executions counters when retrying

Otherwise, we'd retry the job with the exhausted counters from before, if there
was any automatic retry at all, and no more automatic retries will be performed.

Author: rosa

app/models/solid_queue/failed_execution.rb

@@ -21,6 +21,7 @@ def self.retry_all(jobs)
     def retry
       SolidQueue.instrument(:retry, job_id: job.id) do
         with_lock do
+          job.reset_execution_counters
           job.prepare_for_execution
           destroy!
         end

app/models/solid_queue/job/executable.rb

@@ -6,16 +6,14 @@ module Executable
       extend ActiveSupport::Concern
 
       included do
-        include ConcurrencyControls, Schedulable
+        include ConcurrencyControls, Schedulable, Retryable
 
         has_one :ready_execution
         has_one :claimed_execution
-        has_one :failed_execution
 
         after_create :prepare_for_execution
 
         scope :finished, -> { where.not(finished_at: nil) }
-        scope :failed, -> { includes(:failed_execution).where.not(failed_execution: { id: nil }) }
       end
 
       class_methods do
@@ -97,18 +95,10 @@ def status
         end
       end
 
-      def retry
-        failed_execution&.retry
-      end
-
       def discard
         execution&.discard
       end
 
-      def failed_with(exception)
-        FailedExecution.create_or_find_by!(job_id: id, exception: exception)
-      end
-
       private
         def ready
           ReadyExecution.create_or_find_by!(job_id: id)

app/models/solid_queue/job/retryable.rb

@@ -0,0 +1,29 @@
+# frozen_string_literal: true
+
+module SolidQueue
+  class Job
+    module Retryable
+      extend ActiveSupport::Concern
+
+      included do
+        has_one :failed_execution
+
+        scope :failed, -> { includes(:failed_execution).where.not(failed_execution: { id: nil }) }
+      end
+
+      def retry
+        failed_execution&.retry
+      end
+
+      def failed_with(exception)
+        FailedExecution.create_or_find_by!(job_id: id, exception: exception)
+      end
+
+      def reset_execution_counters
+        arguments["executions"] = 0
+        arguments["exception_executions"] = {}
+        save!
+      end
+    end
+  end
+end

test/dummy/app/jobs/raising_job.rb

@@ -7,10 +7,10 @@ class DiscardableError < StandardError; end
   retry_on DefaultError, attempts: 3, wait: 0.1.seconds
   discard_on DiscardableError
 
-  def perform(raising, identifier, attempts = 1)
+  def perform(raising, identifier, fail_count = 1)
     raising = raising.shift if raising.is_a?(Array)
 
-    if raising && executions <= attempts
+    if raising && executions <= fail_count
       JobBuffer.add("#{identifier}: raised #{raising} for the #{executions.ordinalize} time")
       raise raising, "This is a #{raising} exception"
     else

test/integration/jobs_lifecycle_test.rb

@@ -62,6 +62,26 @@ class JobsLifecycleTest < ActiveSupport::TestCase
     assert_equal 1, SolidQueue::FailedExecution.count
   end
 
+  test "retry job that failed after being automatically retried" do
+    RaisingJob.perform_later(RaisingJob::DefaultError, "A", 5)
+
+    @dispatcher.start
+    @worker.start
+
+    wait_for_jobs_to_finish_for(3.seconds)
+
+    assert_equal 2, SolidQueue::Job.finished.count # 2 retries of A
+    assert_equal 1, SolidQueue::FailedExecution.count
+
+    failed_execution = SolidQueue::FailedExecution.last
+    failed_execution.job.retry
+
+    wait_for_jobs_to_finish_for(3.seconds)
+
+    assert_equal 4, SolidQueue::Job.finished.count # Add other 2 retries of A
+    assert_equal 1, SolidQueue::FailedExecution.count
+  end
+
   test "enqueue and run jobs that fail and it's discarded" do
     RaisingJob.perform_later(RaisingJob::DiscardableError, "A")