Fixing potential job-loss race condition

HISTORY.md

@@ -1,6 +1,8 @@
 ##
 
 * Add support for Resque.inline configuration (carlosantoniodasilva)
+* Fixing possible job loss race condition around deleting delayed queues
+  and enqueuing a job 0 seconds in the future.
 
 ## 2.0.0.h (2012-03-19)
 

README.markdown

@@ -6,7 +6,13 @@ resque-scheduler
 Resque-scheduler is an extension to [Resque](http://github.com/defunkt/resque)
 that adds support for queueing items in the future.
 
-Requires redis >=1.3.
+This table explains the version requirements for redis
+
+| resque-scheduler version | required redis version|
+|:-------------------------|----------------------:|
+| >= 2.0.0.i               | >= 2.2.0              |
+| <= 2.0.0.i               | >= 1.3                |
+
 
 Job scheduling is supported in two different way: Recurring (scheduled) and
 Delayed.

lib/resque_scheduler.rb

@@ -258,12 +258,16 @@ def job_to_hash_with_queue(queue, klass, args)
 
     def clean_up_timestamp(key, timestamp)
       # If the list is empty, remove it.
+      redis.watch key
       if 0 == redis.llen(key).to_i
-        redis.del key
-        redis.zrem :delayed_queue_schedule, timestamp.to_i
+        redis.multi do
+          redis.del key
+          redis.zrem :delayed_queue_schedule, timestamp.to_i
+        end
+      else
+        redis.unwatch
       end
     end
-
     def validate_job!(klass)
       if klass.to_s.empty?
         raise Resque::NoClassError.new("Jobs must be given a class.")