MySql - Abysmal Performance - mysql

Question

I am trying to run a relatively simple query on a table that has half a million rows. It's just a small fragment I'm using to test the values I get back are correct. The problem is this query takes over 20 minutes to complete, which seems unusually slow even for 500,000 records.
DROP VIEW IF EXISTS view_temp_sortie_stats;
CREATE VIEW view_temp_sortie_stats AS
SELECT server_id, session_id, ucid, role, sortie_id,
(
SELECT COUNT(sortie_id)
FROM raw_gameevents_log
WHERE sortie_id = l.sortie_id AND server_id = l.server_id AND session_id = l.session_id AND target_player_ucid = l.ucid AND event = "HIT"
) AS HitsReceived
FROM raw_gameevents_log l
WHERE ucid IS NOT NULL AND sortie_id IS NOT NULL
GROUP BY server_id, session_id, ucid, role, sortie_id;
SELECT * FROM view_temp_sortie_stats;
Here is my table:
Next I tried to add indexes for server_id, session_id, sortie_id to see if it would improve - this took more than 10 minutes to apply and timed out. So I could not add them.
This seems abnormally slow, it shouldn't take this much time to add indexes, or perform this query.
My innodb_buffer_pool_size is 5GB, yet the mysqld process only consumes 300mb of memory when these queries are run.
I am running on Windows Server 2012 R2 Standard with 12 GB Ram, 2x Intel Haswell CPU, so I should be seeing much better performance than this from mysql.
There is no one else connected to this instance of MySql and no other operations are occurring.
EDIT - Here is the query explained
Does someone know what might be wrong?
EDIT2 - Partial Fix
After some googling I found out why the add index was taking forever - the original query was still running in the background for over 2hrs. Once I Killed the query the add index took about 30 seconds.
Now when I run the above query it takes 27 seconds - which is a drastic improvement for sure, but that still seems pretty slow for 500,000 records. Here is the new query explain plan:

Answer 1

Your subquery is:
SELECT COUNT(sortie_id)
FROM raw_gameevents_log
WHERE sortie_id = l.sortie_id AND server_id = l.server_id
AND session_id = l.session_id AND target_player_ucid = l.ucid
AND event = "HIT"
and the main query is:
SELECT server_id, session_id, ucid, role, sortie_id, [...]
FROM raw_gameevents_log l
WHERE ucid IS NOT NULL AND sortie_id IS NOT NULL
GROUP BY server_id, session_id, ucid, role, sortie_id;
Let's start from the subquery. The COUNT can count on whatever, so we don't bother with the select fields. The WHERE fields:
WHERE sortie_id = l.sortie_id AND server_id = l.server_id
AND session_id = l.session_id AND target_player_ucid = l.ucid
AND event = "HIT"
You create an index beginning with the constant fields, then the others:
CREATE INDEX subqindex ON raw_gameevents_log(
event,
sortie_id, server_id, session_id, target_player_ucid
)
Then the main query:
WHERE ucid IS NOT NULL AND sortie_id IS NOT NULL
GROUP BY server_id, session_id, ucid, role, sortie_id;
Here you need an index on
ucid, sortie_id, server_id, session_id, role
Finally, you might try getting rid of the subquery (even if the optimizer probably already does a good work with that):
SELECT server_id, session_id, ucid, role, sortie_id,
COALESCE(hits, 0) AS hits
FROM raw_gameevents_log l
LEFT JOIN
(
SELECT COUNT(1) AS hits FROM raw_gameevents_log
WHERE event = 'HIT'
) AS h
ON (h.sortie_id = l.sortie_id, h.server_id = l.server_id, h.session_id = l.session_id, h.target_player_ucid = l.ucid)
WHERE l.ucid IS NOT NULL AND l.sortie_id IS NOT NULL
GROUP BY l.server_id, l.session_id, l.ucid, l.role, l.sortie_id;