MySQL INNODB SELECT query hanging on copying to tmp table

MySQLmysql-5performance

I have a query that is hanging (seemingly) randomly on a specific query. Most of my website users do not have a problem when this query runs but a few users are having timeout issues on the website and when I check the processlist it gets stuck on copying to tmp table.

I believe the hard drive is just not keeping up, which is causing the hang on the copying to tmp table.

I don't believe I can optimize the query any more efficiently. I was wondering if I could set up my server's second hard drive as the temp table storage so that way I'm spreading out some of the load being put on the server.

The query:
SELECT DISTINCT  
    activity_seismo_info.ID, 
    activity_seismo_info.CREATED_AT, 
    activity_seismo_info.UPDATED_AT, 
    ....    
    activity_seismo_info.ASI_EXTRA_5, 
    seismo.ID, 
    ....    
    seismo.SEISMO_NOTES 

FROM `activity_seismo_info` 
    JOIN activity_info ON (activity_seismo_info.ASI_ACTIVITY_ID=activity_info.ID) 
    JOIN location ON (activity_info.ACTIVITY_LOCATION_ID=location.ID) 
    LEFT JOIN seismo ON (activity_seismo_info.ASI_SEISMO_ID=seismo.ID) 

WHERE location.ID='193' 

ORDER BY activity_info.ACTIVITY_EVENT_TIME DESC LIMIT 30

I am on a P4 with 2GB RAM and two hard drives (the second hard drive is not used)

PHP5.2

MySQL 5.0.51a-3ubuntu5.4

symfony 1.2 application with Propel

UPDATE

thanks to @DTest's suggestion, I found my query was returning all the rows in the table before running a limit on the result. But for future reference, is there a way to setup MySQL to use a second hard drive as the 'scratch' drive for temp tables?

Best Answer

In my.cnf use this option

tmpdir=/tmp (this is the default anyway, just choose another folder)

Restart of mysql required

BTW since the location.ID is '193' why even involve the location table ???

Try the following :

SELECT DISTINCT
    activity_seismo_info.ID, 
    activity_seismo_info.CREATED_AT, 
    activity_seismo_info.UPDATED_AT, 
    #....    
    activity_seismo_info.ASI_EXTRA_5, 
    seismo.ID, 
    #....    
    seismo.SEISMO_NOTES

FROM `activity_seismo_info`
    JOIN activity_info ON (activity_seismo_info.ASI_ACTIVITY_ID=activity_info.ID)
    JOIN (SELECT ID from location WHERE ID='193') location ON (activity_info.ACTIVITY_LOCATION_ID=location.ID)
    LEFT JOIN seismo ON (activity_seismo_info.ASI_SEISMO_ID=seismo.ID) 
;

Related Solutions

Mysql – Monstrous query hanging on “copying to tmp table” in MySQL

Wow, that is a doozie of a query. At this point it has little to do with your configuration variables and you should focus on how to optimize the query.

It is tough to give you a magic bullet, though, without knowing what question you are trying to determine from this monstrous query. At first glance, it appears it is trying to do too much.

I would start by removing these two problems from your EXPLAIN plan:

1       PRIMARY <derived2>      ALL     NULL    NULL    NULL    NULL    397520  Using temporary; Using filesort
2       DERIVED <derived3>      ALL     NULL    NULL    NULL    NULL    397520  Using temporary; Using filesort

The first I think is the query for 'order_stats' derived table, the second I am assuming is the 'main' derived table, though it might be flipped. Regardless, I would start at the 'order_stats' (leave a comment with the pastebin of the explain and running time of non-explain, please).

Depending on what this query is trying to accomplish, I would recommend storing the results in a temporary table to be used in the join. One question I would like to know from your developer is, what is the english of the package_tickets and order_tickets computed columns. And if it could be accomplished another way if it turns out to be the bottleneck.

Thesql copying to tmp table duration

Rather than tweaking configs for query optimizing internals, do the following

SUGGESTION #1

You may want to refactor the query as follows

select
    t.id, t.subject,
    count(distinct p.id) as postCount
from thread t left join
(
    select id,thread_id from post where type=0 and 
    posted_on >= DATE_SUB(CURDATE(), INTERVAL 3 DAY)
    order by thread_id
) p on t.id = p.thread_id
group by p.thread_id
order by max(p.posted_on) desc
limit 0, 25;

SUGGESTION #2

Create a compound index for the subquery

ALTER TABLE post
    DROP INDEX type,
    ADD INDEX type_posted_on_ndx (type,posted_on)
;

GIVE IT A TRY !!!

ALTERNATE QUERY

select
    t.id, t.subject,
    count(distinct p.id) as postCount
from thread t left join
(
    select id,thread_id from post where type=0 and 
    posted_on >= DATE_SUB(CURDATE(), INTERVAL 3 DAY)
    order by thread_id
    limit 0,25
) p on t.id = p.thread_id
group by p.thread_id
order by max(p.posted_on) desc;