use correctly named error var

closes FOO-2236

TEST PLAN:
  1) make pulsar writes happen, and then fail
  2) errors that pile up are just the pulsar
     kind, no variable name errors

Change-Id: I167f232b691f78c7a95a9d209147331855332ce7
Reviewed-on: https://gerrit.instructure.com/c/canvas-lms/+/270667
Tested-by: Service Cloud Jenkins <svc.cloudjenkins@instructure.com>
Reviewed-by: Jacob Burroughs <jburroughs@instructure.com>
QA-Review: Ethan Vizitei <evizitei@instructure.com>
Product-Review: Ethan Vizitei <evizitei@instructure.com>

lib/message_bus/async_producer.rb

@@ -162,7 +162,7 @@ module MessageBus
         # context and reconnect.  If we get a timeout again, that is NOT
         # the problem, and we should let the error raise.
         retries += 1
-        raise ex if retries > 1
+        raise e if retries > 1
 
         Rails.logger.info "[AUA] Pulsar failure during message send, retrying..."
         CanvasErrors.capture_exception(:message_bus, e, :warn)

spec/lib/message_bus/async_producer_spec.rb

@@ -21,33 +21,44 @@
 require 'spec_helper'
 
 describe MessageBus::AsyncProducer do
+
   before(:each) do
     skip("pulsar config required to test") unless MessageBus.enabled?
   end
 
-  around(:each) do |example|
-    old_interval = MessageBus.worker_process_interval_lambda
-    old_queue_size = MessageBus.max_mem_queue_size_lambda
-    old_logger = MessageBus.logger
-    # let's not waste time with queue throttling in tests
-    MessageBus.worker_process_interval = -> { 0.01 }
-    MessageBus.max_mem_queue_size = -> { 10 }
-    MessageBus.logger = Rails.logger
-    example.run
-  ensure
-    MessageBus.worker_process_interval = old_interval unless old_interval.nil?
-    MessageBus.max_mem_queue_size_lambda = old_queue_size unless old_queue_size.nil?
-    MessageBus.logger = old_logger
+  describe "#produce_message error handling" do
+    it "re-raises errors correctly if they're not rescuable" do
+      Bundler.require(:pulsar)
+      producer = ::MessageBus::AsyncProducer.new(start_thread: false)
+      allow(MessageBus).to receive(:producer_for).and_raise(::Pulsar::Error::AlreadyClosed)
+      expect{ producer.send(:produce_message, ['a', 'b', 'c']) }.to raise_error(::Pulsar::Error::AlreadyClosed)
+    end
   end
 
-  after(:each) do
-    MessageBus.reset!
-  end
-
-  let(:producer){ MessageBus::AsyncProducer.new(start_thread: false) }
-  let(:namespace){ "test-only" }
-
   describe "push" do
+
+    around(:each) do |example|
+      old_interval = MessageBus.worker_process_interval_lambda
+      old_queue_size = MessageBus.max_mem_queue_size_lambda
+      old_logger = MessageBus.logger
+      # let's not waste time with queue throttling in tests
+      MessageBus.worker_process_interval = -> { 0.01 }
+      MessageBus.max_mem_queue_size = -> { 10 }
+      MessageBus.logger = Rails.logger
+      example.run
+    ensure
+      MessageBus.worker_process_interval = old_interval unless old_interval.nil?
+      MessageBus.max_mem_queue_size_lambda = old_queue_size unless old_queue_size.nil?
+      MessageBus.logger = old_logger
+    end
+
+    after(:each) do
+      MessageBus.reset!
+    end
+
+    let(:producer){ MessageBus::AsyncProducer.new(start_thread: false) }
+    let(:namespace){ "test-only" }
+
     it "pushes onto the queue but does not execute" do
       topic_name = "lazily-created-topic-#{SecureRandom.hex(16)}"
       subscription_name = "subscription-#{SecureRandom.hex(4)}"