Dinghang commented on code in PR #26639:
URL: https://github.com/apache/airflow/pull/26639#discussion_r1071622999


##########
airflow/executors/kubernetes_executor.py:
##########
@@ -62,6 +64,50 @@
 KubernetesWatchType = Tuple[str, str, Optional[str], Dict[str, str], str]
 
 
+def multi_threads_queue_process(
+    queue_size: int,
+    queue_type: str,
+    process_method: Callable,
+    max_threads: int,
+    log: Logger,
+    batch_size: Optional[int] = None,
+) -> None:
+    """
+    Helper method to enable multi-threads for processing queues used with 
kubernetes executor
+    :param queue_size: the size of the queue getting processed
+    :param queue_type: the type of the queue
+    :param process_method: the real method processing the queue
+    :param max_threads: the max num of threads to be used
+    :param log: log
+    :param batch_size: the max num of items we want to process in this round.
+                       If it's not set, the current queue size will be used.
+    """
+    if queue_size == 0:
+        log.info(f'There is no item to process in the {queue_type} queue.')
+        return
+
+    start_time = time.time()
+    log.info(f'Start processing {queue_type} queue with at most {max_threads} 
threads.')
+
+    batch_size = min(batch_size or queue_size, queue_size)
+    max_threads = min(max_threads, queue_size)
+
+    threads = []
+    quotient, remainder = divmod(batch_size, max_threads)
+    for i in range(max_threads):
+        sub_batch_size = quotient + 1 if i < remainder else quotient
+        t = Thread(target=process_method, args=[sub_batch_size])
+        threads.append(t)
+        t.start()
+    for t in threads:
+        t.join()

Review Comment:
   Hi @dstandish  thanks for the info. Most likely would like to keep the 
current implementation. About the "one thread may finish more quickly than the 
others". This could happen but it's not a problem since we are focusing on 
improving is for extreme case with hundreds or thousands items in the queue. 
Overall threads are making it more efficient. 



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to