如何以超過 15 秒的速度對超過 300 萬條記錄的表進行此查詢？

Question

我有以下查詢：

EXPLAIN ANALYZE
SELECT
    customer_id
FROM
    orders 
WHERE
    "status" IN ( 'authorized', 'paid', 'partially_paid', 'pending')    
GROUP BY 
    customer_id
HAVING 
    COUNT(customer_id) >= 2

這會產生以下查詢計划：

Finalize GroupAggregate  (cost=440054.50..516225.43 rows=252557 width=33) (actual time=12206.961..17389.057 rows=457301 loops=1)
  Group Key: customer_id
  Filter: (count(customer_id) >= 2)
  Rows Removed by Filter: 592730
  ->  Gather Merge  (cost=440054.50..511174.29 rows=505114 width=41) (actual time=12206.945..16674.249 rows=1615901 loops=1)
        Workers Planned: 2
        Workers Launched: 2
        ->  Partial GroupAggregate  (cost=439054.47..451871.57 rows=252557 width=41) (actual time=12101.661..14862.466 rows=538634 loops=3)
              Group Key: customer_id
              ->  Sort  (cost=439054.47..442484.98 rows=1372204 width=33) (actual time=12101.648..14344.507 rows=1097122 loops=3)
                    Sort Key: customer_id
                    Sort Method: external merge  Disk: 45448kB
                    ->  Parallel Seq Scan on orders  (cost=0.00..224124.56 rows=1372204 width=33) (actual time=0.014..1205.188 rows=1097122 loops=3)
                          Filter: ((status)::text = ANY ('{authorized,paid,partially_paid,pending}'::text[]))
                          Rows Removed by Filter: 24092
Planning time: 0.101 ms
Execution time: 17434.175 ms

該表本身有超過 300 萬條記錄。

在一天結束的時候，我試圖找到所有已經下過 2 個或更多訂單的客戶，並試圖讓這個查詢快速響應，最好在幾秒鍾內。

我嘗試了幾種方法，但似乎無法讓執行時間變快。

關於如何改進這一點的任何想法？

Answer 1

即使您正在執行順序掃描，掃描實際上也非常快（ actual time=0.014..1205.188毫秒）。 你真正受到打擊的是那種，它需要actual time=12101.648..14344.507 。 您使用Sort Method: external merge Disk: 45448kB

嘗試將work_mem增加到高於 48MB 的值，看看是否有幫助。

Answer 2

我想知道過濾索引是否會加快查詢速度：

create index idf_orders_customer_id_status on (customer_id)
    where "status" IN ( 'authorized', 'paid', 'partially_paid', 'pending') ;

SELECT customer_id
FROM orders 
WHERE "status" IN ( 'authorized', 'paid', 'partially_paid', 'pending')    
GROUP BY customer_id
HAVING COUNT(*) >= 2;