40 seconds strange sql performance puzzle - mysql

Question

I'm running a query to update my user's field like the following:
UPDATE Members SET abc = abc + 1 where Members.id in (
SELECT DISTINCT(memberId) FROM Events WHERE Events.createdAt > "2017-08-06 13:10:00";
Shockingly, with about 500 members, this query runs for 40 seconds...
so the break down:
SELECT DISTINCT(memberId) FROM Events WHERE Events.createdAt > "2017-08-06 13:10:00"
runs for 0.1s and there're only 39 rows matched.
The total # of Members is only ~500. I don't understand why this can take that long... Am I missing something?
I'm running on RDS with mysql 5.6

Answer 1

Try replacing with exists:
UPDATE Members m
SET abc = abc + 1
WHERE EXISTS (SELECT 1
FROM events e
WHERE e.memberId = m.id AND
e.createdAt > '2017-08-06 13:10:00'
);
For performance, you want an index on events(memberId, createdAt).
My guess is that MySQL runs the subquery once for every row in Members. This is consistent with your timing -- ~0.1 seconds * ~500 rows is about 50 seconds, not that far off of 40 seconds.
For SELECTs, this was fixed several versions ago. Perhaps this issue still exists in non-SELECT queries.
You can also write this as:
UPDATE Members m JOIN
(SELECT DISTINCT e.memberId
FROM events e
WHERE e.createdAt > '2017-08-06 13:10:00'
) e
ON e.memberId = m.id
SET abc = abc + 1 ;
Whether this is faster than the exists version, depends on characteristics of your data. Without the suggested index, this is likely to be faster.