SQL长查询优化

Question

Trying to figure out, how to optimize runtime of vital SQL query. Perhaps there is a expert here who could help me with this.

My PHP -> Sql query:

    SELECT ws_manager_repl_log.*, kst_cs_shopping.thread_id, kst_cs_shopping.stage, kst_cs_shopping.user_id
    
    FROM ws_manager_repl_log 
    LEFT JOIN kst_cs_shopping ON ws_manager_repl_log.thread_id = kst_cs_shopping.thread_id
    
    WHERE 
       kst_cs_shopping.stage != 'shoping' 
       AND ws_manager_repl_log.updated_at <= 1599982332
       AND ws_manager_repl_log.mark_as_removed IS NULL
    
    ORDER BY ws_manager_repl_log.updated_at DESC LIMIT 15 OFFSET 15

Here is a problem:

**Run Time: 4.206827**
Select Type Table   Type    Possible Keys   Key Key Len Ref Rows    Extra
SIMPLE  ws_manager_repl_log ALL                 415 Using where; Using temporary; Using filesort
SIMPLE  kst_cs_shopping ALL                 45831   Using where; Using join buffer (flat, BNL join)

Very long runtime... How to iptimize such query? It there any way? Appreciate any help.

Screen 1: https://i.imgur.com/wF0du28.png

Screen 2: Explain: https://i.imgur.com/4lOhyso.png

Screen3: Profiling: https://i.imgur.com/cU2SPgI.png

Screen4: Index def: https://i.imgur.com/UAfioBF.png

Answer 1

First, the LEFT JOIN is superfluous, because the WHERE clause turns it into an INNER JOIN . And table aliases would make the query easier to write and to read:

SELECT mrl.*, s.thread_id, s.stage, s.user_id    
FROM ws_manager_repl_log mrl JOIN
     kst_cs_shopping s
     ON mrl.thread_id = s.thread_id    
WHERE s.stage <> 'shoping' AND
      mrl.updated_at <= 1599982332 AND
      mrl.mark_as_removed IS NULL    
ORDER BY mrl.updated_at DESC
LIMIT 15 OFFSET 15;

It is not clear what indexes would be best for this. One possibility is ws_manager_repl_log(mark_as_removed, updated_at, thread_id) and kst_cs_shopping, thread_id, stage, user_id) .

That said, there are other possibilities, but the question doesn't provide enough information. What filters (if any) are the most selective? What is the relationship between the tables, 1-1? 1-n? n-1? How many rows are returned before the limit ?