Google Cloud Tasks 'Create Task' 请求正在抛出 ServiceUnavailable：503

Question

我正在将我的任务从 AppEngine TaskQueues 转换为 Google Cloud Tasks。

有问题的是每小时检查 S3 存储桶中是否有新文件的 cron 作业。 cron 作业为找到的每个文件启动一个新任务。然后，这些任务会下载各自的文件，并针对文件中的每条记录启动一个新任务。

在此扇出过程中，对 create_task() 的某些调用似乎因 ServiceUnavailable: 503 (https://googleapis.dev/python/cloudtasks/latest/gapic/v2/api.html#google.cloud.tasks_v2.CloudTasksClient.create_task)

而失败

这是一个

Traceback (most recent call last):
  ...
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/src/utils/gc_tasks.py", line 72, in _gc_create_task
    _ = _tasks_client.create_task(parent=_queue_path(DEFAULT_QUEUE), task=task)
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/cloud/tasks_v2/gapic/cloud_tasks_client.py", line 1512, in create_task
    request, retry=retry, timeout=timeout, metadata=metadata
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/api_core/gapic_v1/method.py", line 143, in __call__
    return wrapped_func(*args, **kwargs)
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/api_core/retry.py", line 273, in retry_wrapped_func
    on_error=on_error,
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/api_core/retry.py", line 182, in retry_target
    return target()
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/api_core/timeout.py", line 214, in func_with_timeout
    return func(*args, **kwargs)
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/api_core/grpc_helpers.py", line 59, in error_remapped_callable
    six.raise_from(exceptions.from_grpc_error(exc), exc)
  File "/base/alloc/tmpfs/dynamic_runtimes/python27g/ebb3af67a06047b6/python27/python27_lib/versions/third_party/six-1.12.0/six/__init__.py", line 737, in raise_from
    raise value
ServiceUnavailable: 503 {
    "created":"@1583436423.131570193",
    "description":"Delayed close due to in-progress write",
    "file":"third_party/apphosting/python/grpcio/v1_0_0/src/core/ext/transport/chttp2/transport/chttp2_transport.c",
    "file_line":412,
    "grpc_status":14,
    "referenced_errors":[{
        "created":"@1583436423.131561040",
        "description":"OS Error",
        "errno":32,
        "file":"third_party/apphosting/python/grpcio/v1_0_0/src/core/lib/iomgr/tcp_posix.c",
        "file_line":393,
        "os_error":"Broken pipe",
        "syscall":"sendmsg"}
    ]}

这是另一个

Traceback (most recent call last):
  ...
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/src/utils/pt_gc_tasks.py", line 72, in _gc_create_task
    _ = _tasks_client.create_task(parent=_queue_path(DEFAULT_QUEUE), task=task)
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/cloud/tasks_v2/gapic/cloud_tasks_client.py", line 1512, in create_task
    request, retry=retry, timeout=timeout, metadata=metadata
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/api_core/gapic_v1/method.py", line 143, in __call__
    return wrapped_func(*args, **kwargs)
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/api_core/retry.py", line 273, in retry_wrapped_func
    on_error=on_error,
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/api_core/retry.py", line 182, in retry_target
    return target()
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/api_core/timeout.py", line 214, in func_with_timeout
    return func(*args, **kwargs)
  File "/base/data/home/apps/s~my_project/dev.XXXXXXXXXXXXXXXXXXX/lib/google/api_core/grpc_helpers.py", line 59, in error_remapped_callable
    six.raise_from(exceptions.from_grpc_error(exc), exc)
  File "/base/alloc/tmpfs/dynamic_runtimes/python27g/ebb3af67a06047b6/python27/python27_lib/versions/third_party/six-1.12.0/six/__init__.py", line 737, in raise_from
    raise value
ServiceUnavailable: 503 {
    "created":"@1583407622.505288938",
    "description":"Endpoint read failed",
    "file":"third_party/apphosting/python/grpcio/v1_0_0/src/core/ext/transport/chttp2/transport/chttp2_transport.c",
    "file_line":1807,
    "grpc_status":14,
    "occurred_during_write":0,
    "referenced_errors":[{
        "created":"@1583407622.505108366",
        "description":"Secure read failed",
        "file":"third_party/apphosting/python/grpcio/v1_0_0/src/core/lib/security/transport/secure_endpoint.c",
        "file_line":158,
        "referenced_errors":[{
            "created":"@1583407622.505106550",
            "description":"Socket closed",
            "file":"third_party/apphosting/python/grpcio/v1_0_0/src/core/lib/iomgr/tcp_posix.c",
            "file_line":259}
        ]}
    ]}

我是否同时排队太多任务？我能做些什么来解决这个问题？

Answer 1

您分享的两个错误的原因似乎与其描述中的文本不同，但两者确实都可能与您队列中的任务过载有关。

您可以采取的解决方法是设置一些速率限制以降低负载，或者您可以设置重试参数，因为显然它只发生在少数任务中。无论您选择哪种方式，您都可以在 Cloud Task Configuring Queue Documentation.

中找到方法

Answer 2

HTTP Error 503. The service is unavailable 是对应的Wep Application的Application Pool被Stopped或Disabled或Paused时出现的。或由于密码过期或锁定，应用程序池的给定用户身份可能无效。

Answer 3

经过大量挖掘，“503 服务不可用”似乎是所有 GCP 服务的 google-cloud SDK 中一个非常常见的错误。

Why do I get 503 Service Unavailable errors using the Google Cloud Datastore API explorer?
https://github.com/googleapis/google-cloud-python/issues/3128

解决方案是启用重试逻辑。 google-cloud-core（google-cloud-tasks 依赖）有一个现有的重试机制，但没有为任务创建配置。

retry_codes_name 被设置为 non_idempotent 而不是 idempotent

            "CreateTask": {
                "timeout_millis": 10000,
                "retry_codes_name": "non_idempotent",
                "retry_params_name": "default",
            },

https://github.com/googleapis/python-tasks/blob/100e9c709383848498e1e6a747fb819520a7d8c1/google/cloud/tasks_v2/gapic/cloud_tasks_client_config.py#L87

我的猜测是这可能会导致重复的任务排队。但是，如果您指定了一个任务名称，google-cloud-tasks 应该可以防止这些重复项被排队。

所以我将一个 Retry 对象传递给 .create_task() 而没有为 predicate 提供 arg，这导致它默认为 if_transient_error() 它将重试以下错误: exceptions.InternalServerError, exceptions.TooManyRequests, exceptions.ServiceUnavailable

下面是我创建任务的代码片段

from google.api_core import retry
from google.api_core.exceptions import AlreadyExists
from google.cloud import tasks

_tasks_client = tasks.CloudTasksClient()


def my_create_task_function(my_queue_path, task_object):
    try:
        _tasks_client.create_task(
            parent=my_queue_path, 
            task=task_object, 
            retry=retry.Retry(  # Copies the default retry config from retry_params in google.cloud.tasks_v2.gapic.cloud_tasks_client_config
                initial=.1,
                maximum=60,
                multiplier=1.3,
                deadline=600))
    except AlreadyExists:
        logging.warn("found existing task")

还有一个可用的记录器，您可以调整其级别，以便您可以查看实际重试时的日志语句。

如果您执行以下操作：

logging.getLogger('google.api_core.retry').setLevel(logging.DEBUG)

当它启动时，您应该会在日志中看到这样的消息：

Google Cloud Tasks 'Create Task' 请求正在抛出 ServiceUnavailable：503

Google Cloud Tasks 'Create Task' request is throwing ServiceUnavailable: 503

google-app-engine

google-cloud-python

google-cloud-tasks