Question

在添加这两行或其中任何一行后，查询超时

AND final IS NOT NULL 
ORDER BY tmodified ASC

查询持续运行超过10分钟......并且超时。

如果我删除上面两行，它会在1毫秒内返回结果，这可以正常工作。

任何想法如何使以上两行使用以下查询？

表table_h有3600万条记录，在此表中

column final is  numeric(10,5)
column tmodified is bigint, Timestamp

我正在使用Postgres 9.4

这是完整的查询。

SELECT DISTINCT t.cid, h.a, am.b, u2.c, u.d, h.e, ie.f, im.g FROM table_am as am
  INNER JOIN table_t as t on (t.id = am.id AND t.type = am.type)
  INNER JOIN table_h AS h on h.iid = t.id
  INNER JOIN table_u AS u ON u.id = h.uid
  INNER JOIN table_u AS u2 ON u2.id = h.lu
  INNER JOIN table_im AS im ON im.asid = am.sid
  INNER JOIN table_ie AS ie ON ie.uid = u.uid
  INNER JOIN table_g AS g ON g.id = h.oldid
WHERE h.final >= 0 
  AND h.final IS NOT NULL
  AND h.tmodified >= 1499903419
  AND ie.p = im.p
  AND h.sr IN ('x', 'y', 'z')
  AND h.id = (SELECT id FROM table_h WHERE oldid = h.oldid AND final >= 0 
              AND final IS NOT NULL -- Issue is here and
              ORDER BY tmodified ASC -- Issue is here
              LIMIT 1)
  AND h.id NOT IN (SELECT id FROM table_m  WHERE tmodified > 1499903419)

Answer 1

尝试使用连接替换相关的子查询，如下所示：

SELECT ...
FROM table_am as am
...
INNER JOIN table_g AS g ON g.id = h.oldid
INNER JOIN
(
    SELECT id, oldid, MIN(tmodified) AS min_mod
    FROM table_h
    WHERE final >= 0 AND final IS NOT NULL
    GROUP BY id, oldid
) t
    ON h.id     = t.id    AND
       h.oldid  = t.oldid AND
       h.tmodified = t.min_mo

请注意，我们也可以使用行号或其他分析函数表达这一点，但希望这为您提供了一个开始的地方。

Answer 2

好吧，我可以解决你的一半问题。条件：

AND h.final IS NOT NULL

不需要。条件：

h.final >= 0

已考虑到这一点。

如果剩余查询返回的速度很快，则使用子查询或cte，然后按：

排序

with cte as (
      select . . ., t.modified
     )
select cte.*
from cte
order by modified;

Answer 3

吸尘所有涉及表将解决问题。

VACUUM ANALYZE table_am;
VACUUM ANALYZE table_t;
VACUUM ANALYZE table_h;
VACUUM ANALYZE table_u;
VACUUM ANALYZE table_im;
VACUUM ANALYZE table_ie;
VACUUM ANALYZE table_g;
VACUUM ANALYZE table_m;

参考：https://www.postgresql.org/docs/9.4/static/sql-vacuum.html

Postgres 9.4 sql查询超时

3 个答案: