appengine，python：taskqueue.add() 中是否存在内存泄漏？

Question

以下代码添加了对 blobstore 中的文件执行某些处理的任务，它在 B2 后端上运行，因此没有超时限制：

for task in tasks:
    tools.debug("add_tasks_to_process_files", "adding_task")

    taskqueue.add(\
            name=("Process_%s_files---%s--%s--%s--%s" % \
                        (len(tasks[task]), task[1], task[0], task[2], int(time.time()))),\
            queue_name="files-processor",\
            url="/analytics/process_files/",\
            params={"processing_task": json.dumps({"profile": task, "blobs_to_process": tasks[task]})})

任务是以下形式的字典：

{
    (x1,y1,z1): ["blob_key", "blob_key"... (limited to 35 keys)],
    (x2,y2,z2): ["blob_key", "blob_key"...],
    .
    .
    .
}

x1、y1、z1 是所有字符串

tools.debug是我编写的一个函数，它使用urlfetch将消息发送到我的本地服务器（这样我就不必等待20分钟才能读取日志）：

def debug(location, message, params=None, force=False):
    if not (settings.REMOTE_DEBUG or settings.LOCALE_DEBUG or force):
        return

    if params is None:
        params = {}

    params["memory"] = runtime.memory_usage().current()
    params["instance_id"] = settings.INSTANCE_ID

    debug_message = "%s/%s?%s" % (urllib2.quote(location), urllib2.quote(message), "&".join(["%s=%s" % (p, urllib2.quote(unicode(params[p]).encode("utf-8"))) for p in params]))

    if settings.REMOTE_DEBUG or force:
        fetch("%s/%s" % (settings.REMOTE_DEBUGGER, debug_message))

    if settings.LOCALE_DEBUG or force:
        logging.debug(debug_message)

因为tools.debug不在代码中第一次失败，我确信这不是内存问题的原因。

我收到此错误：

   /add_tasks_to_process_files/ 500 98812ms 0kb instance=0 AppEngine-Google; (+http://code.google.com/appengine):
    A serious problem was encountered with the process that handled this request, causing it to exit. This is likely to cause a new process to be used for the next request to your application. If you see this message frequently, you may have a memory leak in your application. (Error code 201)

就在它之后：

/_ah/stop 500 110ms 0kb
Exceeded soft private memory limit with 283.406 MB after servicing 1 requests total

再次，我收到了上面的代码，但没有行： tools.debug("add_tasks_to_process_files", "adding_task")

现在，让我向您展示我的内容在我的调试器中查看：

1 2012-1-19 14:41:38 [processors-backend] processors-backend-initiated instance_id: 1329662498, memory: 18.05078125, backend_instance_url: http://0.processors.razoss-dock-dev.appspot.com, backend_load_balancer_url: http://processors.razoss-dock-dev.appspot.com
2 2012-1-19 14:41:39 [AddTasksToProcessFiles] start instance_id: 1329662498, files_sent_to_processing_already_in_previous_failed_attempts: 0, memory: 19.3828125
3 2012-1-19 14:41:59 [AddTasksToProcessFiles] add_tasks_to_process_files-LOOP_END total_tasks_to_add: 9180, total_files_added_to_tasks: 9184, task_monitor.files_sent_to_processing: 0, total_files_on_tasks_dict: 9184, instance_id: 1329662498, memory: 56.52734375
4 2012-1-19 14:42:0 [add_tasks_to_process_files] adding_task instance_id: 1329662498, memory: 57.81640625
5 2012-1-19 14:42:0 [add_tasks_to_process_files] adding_task instance_id: 1329662498, memory: 57.81640625
6 2012-1-19 14:42:1 [add_tasks_to_process_files] adding_task instance_id: 1329662498, memory: 57.9375
7 2012-1-19 14:42:2 [add_tasks_to_process_files] adding_task instance_id: 1329662498, memory: 57.9375
8 2012-1-19 14:42:2 [add_tasks_to_process_files] adding_task instance_id: 1329662498, memory: 58.03125
.
.
.
2183 2012-1-19 14:53:45 [add_tasks_to_process_files] adding_task instance_id: 1329662498, memory: 280.66015625
2184 2012-1-19 14:53:45 [add_tasks_to_process_files] adding_task instance_id: 1329662498, memory: 280.66015625
2185 2012-1-19 14:53:45 [add_tasks_to_process_files] adding_task instance_id: 1329662498, memory: 281.0
2    186 2012-1-19 14:53:46 [add_tasks_to_process_files] adding_task instance_id: 1329662498, memory: 281.0
2187 2012-1-19 14:53:46 [add_tasks_to_process_files] adding_task instance_id: 1329662498, memory: 281.0
2188 2012-1-19 14:53:46 [add_tasks_to_process_files] adding_task instance_id: 1329662498, memory: 281.3828125

完整跟踪： http://pastebin.com/CcPDU6s7

taskqueue.add() 中是否存在内存泄漏？

谢谢

Answer 1

虽然这不能回答您的特定问题，但您是否尝试过使用 Queue 批量添加任务？

http://code.google.com/appengine/docs/python /taskqueue/queues.html#Queue_add

您一次最多可以添加 100 个任务。

http://code.google.com/appengine/docs /python/taskqueue/overview-push.html#Quotas_and_Limits_for_Push_Queues

未经测试的代码。

queue = taskqueue.Queue(name="files-processor")
while tasks:
    queue.add(taskqueue.Task(...) for k,v in (tasks.popitem() for _ in range(min(len(tasks),100))))

如果您仍然想在其他地方使用任务，则必须稍微更改此结构（或制作副本）。