Modified the task locking documentation to use cache.add for atomic locking

docs/cookbook/tasks.rst

@@ -40,23 +40,21 @@ The cache key expires after some time in case something unexpected happens
             feed_url_digest = md5(feed_url).hexdigest()
             lock_id = "%s-lock-%s" % (self.name, feed_url_hexdigest)
 
-            is_locked = lambda: str(cache.get(lock_id)) == "true"
-            acquire_lock = lambda: cache.set(lock_id, "true", LOCK_EXPIRE)
-            # memcache delete is very slow, so we'd rather set a false value
-            # with a very low expiry time.
-            release_lock = lambda: cache.set(lock_id, "nil", 1)
+            # cache.add fails if if the key already exists
+            acquire_lock = lambda: cache.add(lock_id, "true", LOCK_EXPIRE)
+            # memcache delete is very slow, but we have to use it to take
+            # advantage of using add() for atomic locking
+            release_lock = lambda: cache.delete(lock_id)
 
             logger.debug("Importing feed: %s" % feed_url)
-            if is_locked():
-                logger.debug(
-                    "Feed %s is already being imported by another worker" % (
-                        feed_url))
-                return
-
-            acquire_lock()
-            try:
-                feed = Feed.objects.import_feed(feed_url)
-            finally:
-                release_lock()
-
-            return feed.url
+            if aquire_lock():
+                try:
+                    feed = Feed.objects.import_feed(feed_url)
+                finally:
+                    release_lock()
+                return feed.url
+
+            logger.debug(
+                "Feed %s is already being imported by another worker" % (
+                    feed_url))
+            return