sql query optimization sugarcrm - mysql

Question

I am trying to optimize sql query in mysql db. Tried various variations with indexes, but nothing helps. Maybe I am missing something
Query:
SELECT count(1) AS fAccounts
from sugarcrm.accounts t4,
( SELECT t3.related_id
FROM sugarcrm.prospect_lists_prospects t3, sugarcrm.prospect_list_campaigns t2
where t3.deleted=0
and t3.related_type='Accounts'
and t3.prospect_list_id=t2.prospect_list_id
and t2.deleted=0
and t2.campaign_id='10909eb7-8080-45b6-8c9f-563b42be91e5'
) t3
where t4.deleted=0
and t4.id=t3.related_id;
Explain:
+----+-------------+------------+--------+---------------------------------------------------+----------------+---------+------------------------------+--------+-------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+------------+--------+---------------------------------------------------+----------------+---------+------------------------------+--------+-------------+
| 1 | PRIMARY | <derived2> | ALL | NULL | NULL | NULL | NULL | 5000 | |
| 1 | PRIMARY | t4 | eq_ref | PRIMARY;idx_accnt_id_del;idx_accnt_assigned_del | PRIMARY | 108 | t3.related_id | 1 | Using where |
| 2 | DERIVED | t2 | ref | idx_pro_id;idx_cam_id;idx_prospect_list_campaigns | idx_cam_id | 111 | | 1 | Using where |
| 2 | DERIVED | t3 | ref | idx_plp_pro_id;idx_plp_rel_id_2 | idx_plp_pro_id | 111 | sugarcrm.t2.prospect_list_id | 463968 | Using where |
+----+-------------+------------+--------+---------------------------------------------------+----------------+---------+------------------------------+--------+-------------+

Answer 1

The inner query is the trouble maker. There are two ways it could be performed: Start with t2 and do a "Nested Loop Join" to t3 or vice versa. The Optimizer will look at the WHERE clause and the table sizes and the indexes to estimate which one would be best to start with. Let's give the optimizer the 'best' index for going each way:
Starting with t2:
t2: INDEX(deleted, campaign_id) -- in either order
t3: INDEX(prospect_list_id, deleted, related_type) -- in any order
Starting with t3:
t3: INDEX(deleted, related_type) -- in either order
t2: INDEX(prospect_list_id, deleted, campaign_id) -- in any order
Rather than adding 2 indexes to each table, let's do
t2: INDEX(campaign_id, deleted, prospect_list_id) -- in this order
t3: INDEX(related_type, deleted, prospect_list_id) -- in this order
Similarly, t4 (which will be last) needs
INDEX(deleted, id)
unless it is InnoDB and already has PRIMARY KEY(id), which will be 'clustered' with the data.
There is a problem... When you do a JOIN, then compute aggregates, the JOIN first gives you an explosion of rows, then the COUNT() counts too many of them, thereby getting an inflated number. So, be sure to sanity check the results.
Since the only need for t4 is to verify that the related_id is there, the query could be reformulated as
SELECT COUNT(*) AS fAccounts
FROM prospect_lists_prospects t3
-- Note the use of `JOIN...ON...`:
JOIN prospect_list_campaigns t2 ON t3.prospect_list_id=t2.prospect_list_id
where t3.deleted=0
and t3.related_type='Accounts'
and t2.deleted=0
and t2.campaign_id='10909eb7-8080-45b6-8c9f-563b42be91e5'
AND ( EXISTS *
FROM accounts
FROM accounts t4
WHERE t4.id = t3.related_id
)
This still needs the suggested indexes (one per table).

Answer 2

Since you don't use DISTINCT anywhere, I see no need to bother creating a temporary table. Try this one:
SELECT
count(1) AS fAccounts
from sugarcrm.accounts t4 inner join
sugarcrm.prospect_lists_prospects t3 on t4.id=t3.related_id inner join
sugarcrm.prospect_list_campaigns t2 on t3.prospect_list_id=t2.prospect_list_id
where t3.deleted=0
and t3.related_type='Accounts'
and t2.deleted=0
and t2.campaign_id='10909eb7-8080-45b6-8c9f-563b42be91e5'
and t4.deleted=0