Google BigQuery：有时 TABLE_DATE_RANGE 不稳定

Question

有时在 Google BigQuery 中使用 TABLE_DATE_RANGE 的查询会失败。
但是目标表已经存在，并且失败的查询通过重试成功。

具体例子如下。

---------------------------------------
The Target Table
---------------------------------------
[ Dataset ID ]    my_dataset
[ Table ID ]      my_table_20170519
[ Creation Time ] 2017-05-20 02:00:52

---------------------------------------
The Executed Query
---------------------------------------
SELECT
  column1, column2, .....
FROM
  TABLE_DATE_RANGE(
    my_dataset.my_table_,
    TIMESTAMP('20170519'),
    TIMESTAMP('20170519')
  )
;

---------------------------------------
The 1st Execution and Result
---------------------------------------
[ Job Start Time ]  2017-05-20 02:00:57.513
[ Job End Time ]    2017-05-20 02:00:57.513
[ Result ]          Failure by "FROM clause with table wildcards matches no table"

---------------------------------------
The 2nd Execution and Result (Retry)
---------------------------------------
[ Job Start Time ]  2017-05-20 02:04:56.556
[ Job End Time ]    2017-05-20 02:04:56.556
[ Result ]          Failure by "FROM clause with table wildcards matches no table"

---------------------------------------
The 3rd Execution and Result (Retry)
---------------------------------------
[ Job Start Time ]  2017-05-20 02:06:43.937
[ Job End Time ]    2017-05-20 02:06:46.291
[ Result ]          Success
-

顺便说一句，使用 NO TABLE_DATE_RANGE 的查询总是会成功。
（例如FROM [my_dataset.my_table_20170519]）。
使用相同的 FROM 子句几乎同时执行的其他查询有时会成功。

当然，使用重试可以解决上面的例子。
但我担心 my_table_20170519 会被使用以下 FROM 子句的查询忽略。

FROM
  TABLE_DATE_RANGE(
    my_dataset.my_table_,
    TIMESTAMP('20170510'),
    TIMESTAMP('20170519')
  )

有没有人有办法解决它？

---

我添加两个关于这个问题的例子。

示例 A：
使用相同的 FROM 子句几乎同时执行的其他查询有时会成功。

----------------------------------------------
The Detail of Case
----------------------------------------------

- A program executes query "A" and query "B" in BigQuery.
- Both queries have the following FROM clause。

  FROM TABLE_DATE_RANGE(
    my_dataset.my_table_,
    TIMESTAMP('20170519'),
    TIMESTAMP('20170519')
  )

----------------------------------------------
The 1st Execution and Result
----------------------------------------------

Query "A"
  [ Job Start Time ]  2017-05-20 02:00:57.513
  [ Job End Time ]    2017-05-20 02:00:57.513
  [ Result ]          Failure by "FROM clause with table wildcards matches no table"

Query "B"
  [ Job Start Time ]  2017-05-20 02:00:57.507
  [ Job End Time ]    2017-05-20 02:01:09.537
  [ Result ]          Success


----------------------------------------------
The 2nd Execution and Result (Retry)
----------------------------------------------

Query "A"
  [ Job Start Time ]  2017-05-20 02:04:56.556
  [ Job End Time ]    2017-05-20 02:04:56.556
  [ Result ]          Failure by "FROM clause with table wildcards matches no table"

# Query "B" is NOT executed because it already succeeded.

----------------------------------------------
The 3rd Execution and Result (Retry)
----------------------------------------------

Query "A"
  [ Job Start Time ]  2017-05-20 02:06:43.937
  [ Job End Time ]    2017-05-20 02:06:46.291
  [ Result ]          Success

# Query "B" is NOT executed because it already succeeded.

示例 B：
此问题有时会在“不小的”时间段后发生。

---------------------------------------
The Target Table
---------------------------------------
[ Dataset ID ]    my_dataset
[ Table ID ]      my_table_b_20170519
[ Creation Time ] 2017-05-20 01:42:22

---------------------------------------
The Executed Query
---------------------------------------
SELECT
  column1, column2, .....
FROM
  TABLE_DATE_RANGE(
    my_dataset.my_table_b_,
    TIMESTAMP('20170519'),
    TIMESTAMP('20170519')
  )
;

----------------------------------------------
The 1st Execution and Result
----------------------------------------------
[ Job Start Time ]  2017-05-20 01:59:51.255
[ Job End Time ]    2017-05-20 01:59:51.255
[ Result ]          Failure by "FROM clause with table wildcards matches no table"

----------------------------------------------
The 2nd Execution and Result
----------------------------------------------
[ Job Start Time ]  2017-05-20 02:04:53.802
[ Job End Time ]    2017-05-20 02:04:57.684
[ Result ]          Success

Answer 1

我认为 TABLE_DATE_RANGE 正在使用元表 (tables.list) 来获取符合条件的表列表。
问题在于，虽然表数据可立即用于查询 - 元数据最终可用 - 这意味着将新表放入列表需要时间。
如果您在创建后立即查询该表，您将不会看到此问题。

注意：以上假设您仅在新表创建后的“小”时间段后才遇到此问题

另见 - Eventually consistent operations - 我觉得它是相关的