Do not connect to redis at exit if not needed

A fix for #4498 introduced
a call-back that is executed at ruby VM exit and tries to connect to redis.

This call-back is also executed in CI runs and in all other cases where sidekiq is loaded but not used with real redis.

As a work-around, keep the at_exit hook but communicate to redis only if there is something to send, i.e. the number of processed or failed jobs is greater than zero

Changes.md

@@ -5,6 +5,7 @@
 HEAD
 ---------
 
+- Fix: Do not connect to redis at ruby vm exit when not needed. [#4502] 
 - Remove Redis connection naming [#4479]
 
 6.0.6

lib/sidekiq/launcher.rb

@@ -100,15 +100,17 @@ def self.flush_stats
       nowdate = Time.now.utc.strftime("%Y-%m-%d")
       fails = Processor::FAILURE.reset
       procd = Processor::PROCESSED.reset
-      Sidekiq.redis do |conn|
-        conn.pipelined do
-          conn.incrby("stat:processed", procd)
-          conn.incrby("stat:processed:#{nowdate}", procd)
-          conn.expire("stat:processed:#{nowdate}", STATS_TTL)
+      if fails + procd > 0
+        Sidekiq.redis do |conn|
+          conn.pipelined do
+            conn.incrby("stat:processed", procd)
+            conn.incrby("stat:processed:#{nowdate}", procd)
+            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.incrby("stat:failed", fails)
+            conn.incrby("stat:failed:#{nowdate}", fails)
+            conn.expire("stat:failed:#{nowdate}", STATS_TTL)
+          end
         end
       end
     end