更新 FROM 子句中使用的同一个表上的记录

Question

我正在尝试更新 FROM 子句中使用的表上的记录。当我执行它时，它需要更多的时间来执行（将近5小时）。 下面附上详细信息。

更新声明

UPDATE schema_name.tbl_bllng upd
SET
rsv_clm = Mc.c_fullform
, counter = tbll.counter + 1
 
FROM
    schema_name.tbl_bllng tbll
    INNER JOIN schema_name.tbl_ihdr ihdr
    ON(
    ihdr.b_id = tbll.b_id
    AND ihdr.Icnt = 0
    AND ihdr.id = 1
    AND ihdr.active_flg = False)
    LEFT OUTER JOIN public.schema_name2.master_table Mc
    ON(
    ihdr.rsv_clm IS NOT NULL
    AND Mc.ic_id = ihdr.rsv_clm
    AND Mc.id = 1
    AND Mc.cd = 1
    AND Mc.active_flg = False)
    WHERE
    tbll.tseq_id = upd.tseq_id
    AND tbll.id = 1
    AND tbll.active_flg = False;

QUERY PLAN
Update on tbl_bllng upd  (cost=1.14..932050009.01 rows=53247247974 width=8053)
  ->  Merge Join  (cost=1.14..932050009.01 rows=53247247974 width=8053)
        Merge Cond: ((tbll.b_id)::text = (ihdr.b_id)::text)
        ->  Nested Loop  (cost=0.71..108174.20 rows=171730 width=7628)
              ->  Index Scan using idx_bllng_id on tbl_bllng tbll  (cost=0.29..13970.46 rows=171730 width=23)
                    Index Cond: ((id = 1) AND (active_flg = false))
              ->  Index Scan using bllng_pkc on tbl_bllng upd  (cost=0.42..0.55 rows=1 width=7613)
                    Index Cond: (tseq_id = tbll.tseq_id)
        ->  Materialize  (cost=0.43..117666.47 rows=1240212 width=95)
              ->  Nested Loop Left Join  (cost=0.43..114565.94 rows=1240212 width=95)
                    Join Filter: ((ihdr.rsv_clm IS NOT NULL) AND ((Mc.ic_id)::text = (ihdr.rsv_clm)::text))
                    ->  Index Scan using idx_tbl_bll on tbl_ihdr ihdr  (cost=0.43..95961.70 rows=1240212 width=429)
                          Index Cond: ((Icnt = 0) AND (id = 1) AND (active_flg = false))
                    ->  Materialize  (cost=0.00..1.06 rows=1 width=162)
                          ->  Seq Scan on "schema_name2.master_table" Mc  (cost=0.00..1.06 rows=1 width=162)
                                Filter: ((NOT active_flg) AND (id = 1) AND (cd = 1))

Table names and Record counts
schema_name2.master_table this table 4 record only
schema_name.tbl_bllng - 171010
schema_name.tbl_ihdr -1240212

Table datatype:
tbl_ihdr
b_id-----(string)
Icnt---integer
rsv_clm --> string
id-----int
active_flg---Boolean

tbl_bllng
b_id--------> string
tseq_id  -- Int
id---int
active_flg---boolean

我是查询调优的新手，请帮助解决此问题。

Answer 1

您不小心进行了

schema_name.tbl_bllng

与自身的自连接。它已经隐式加入为

update

target，因此您无需再次将其列为更新 source。

将指定目标如何与第一个源关联的连接条件下移至

where

部分。您可以保留其他连接不变：

UPDATE schema_name.tbl_bllng upd
SET
rsv_clm = Mc.c_fullform
, counter = upd.counter + 1
FROM schema_name.tbl_ihdr ihdr
    LEFT OUTER JOIN public.schema_name2.master_table Mc
    ON(
    ihdr.rsv_clm IS NOT NULL
    AND Mc.ic_id = ihdr.rsv_clm
    AND Mc.id = 1
    AND Mc.cd = 1
    AND Mc.active_flg = False)
WHERE 
    (ihdr.b_id = upd.b_id
     AND ihdr.Icnt = 0
     AND ihdr.id = 1
     AND ihdr.active_flg = False)
AND upd.tseq_id = upd.tseq_id
    AND upd.id = 1
    AND upd.active_flg = False;