使用索引对MYSQL进行查询优化查询

Question

I have this query

SELECT COUNT(*) as user_count, `partner` FROM (`my_table`) WHERE date_join <='2014-02-19 23:59:59' AND isConfirm=1 GROUP BY `partner`

The table already have 420000 rows and this query is timing out. I am getting the error lost connection to mysql server during query.

I already have indexes for date_join and the explain result is as follows :

id  select_type     table    type   possible_keys   key      key_len    ref     rows    Extra 
1 SIMPLE           my_table  range  date_join     date_join    9        NULL    112223  Using where; Using temporary; Using filesort 

So this seems utilizing the index, but still I am getting the timeout error. How can I optimize further ?

Answer 1

You could try creating an index on both date_join and isConfirm. Example:

alter table `my_table` add index `my_index` (`date_join`,`isConfirm`);

Answer 2

If you are getting approx same rows by below 2 queries apporox. 112223 rows as your explain suggesting, then there is no benefit of creating combined index as suggested by @RKG but if you are getting much less rows then it can be beneficial.

SELECT COUNT(*) FROM my_table WHERE date_join <='2014-02-19 23:59:59';

SELECT COUNT(*) FROM my_table WHERE date_join <='2014-02-19 23:59:59' and isConfirm=1;

If you are getting approx. same rows from both queries then you can check below things to take care this issue:

First check what is set below variables in your config file /etc/my.cnf

tmp_table_size = ???

max_heap_table_size = ???

sort_buffer_size = ???

join_buffer_size = ???

increasing these variables you can get solution but you have to check what total memory in your system and what value will be sufficient for you. As a try you can set values as per below, if your values are less than these values-

tmp_table_size = 1G

max_heap_table_size = 1G

sort_buffer_size = 10M

join_buffer_size = 2M

Note: Make sure your system /tmp directory also should have more than 1 GB space.

---

Even you should add a lower date filter also in your query as you are fetching almost all data till yesterday and then applying group by.

If you need complete count till yesterday then you can try below query. I am not sure if it help you but you can try.

SELECT COUNT(primary_key) AS user_count, `partner` FROM (SELECT primary_key, partner FROM my_table WHERE date_join <='2014-02-19 23:59:59' AND isConfirm=1) a GROUP BY partner;