I have the following problem:

PLEASE EXPLAIN SELECT ‘guys commenting above that they need to pay for support, because you are not fed by International Red Cross database department and the cheapest way to pay is buying your book. If that is too much for them, they should at least buy some bananas, rent some monkeys and start sorting their 20M rows table on plain paper.’;

For the same query”EXPLAIN SELECT avg(length(val)) FROM idxtest WHERE i1=50 AND i2=50;”, why the explain result is different after you set i1=i2.
I think mysql should still choose multi col index not index merge.

For the same query ” EXPLAIN SELECT avg(length(val)) FROM idxtest WHERE i1=50 AND i2=50;” , why there is different explain result after execute query “UPDATE idxtest SET i2=i1;”.
I was confused.

1.
mysql [localhost] {msandbox} (test)> EXPLAIN SELECT avg(length(val)) FROM idxtest WHERE i1=50 AND i2=50;
2.
+—-+————-+———+——+—————-+———-+———+————-+——+——-+
3.
| id | select_type | TABLE | type | possible_keys | KEY | key_len | ref | rows | Extra |
4.
+—-+————-+———+——+—————-+———-+———+————-+——+——-+
5.
| 1 | SIMPLE | idxtest | ref | i1,i2,combined | combined | 8 | const,const | 665 | |
6.
+—-+————-+———+——+—————-+———-+———+————-+——+——-+
7.
1 row IN SET (0.00 sec)

1.
mysql [localhost] {msandbox} (test)> EXPLAIN SELECT avg(length(val)) FROM idxtest WHERE i1=50 AND i2=50;
2.
+—-+————-+———+————-+—————-+——-+———+——+——+————————————-+
3.
| id | select_type | TABLE | type | possible_keys | KEY | key_len | ref | rows | Extra |
4.
+—-+————-+———+————-+—————-+——-+———+——+——+————————————-+
5.
| 1 | SIMPLE | idxtest | index_merge | i1,i2,combined | i2,i1 | 4,4 | NULL | 959 | USING intersect(i2,i1); USING WHERE |
6.
+—-+————-+———+————-+—————-+——-+———+——+——+————————————-+
7.
1 row IN SET (0.00 sec)

I have a scenario here on multiple column keys. I created an partial index for person table of 2M+ records:

CREATE INDEX firstlastname_idx ON person(firstname(2), lastname(2));

The requirement is for wildcard searching on the first character on the firstname and lastname.

select firstname, lastname from person where firstname like ‘p%’ and lastname like ‘s%’;

It’s working fine but I also notice that when the server is idle for some time and then execute a query it takes more than a minute to finished the query. All subsequent query is a snap. So the problem is on the first query execution.

I do run explain and it using the partial index. I know there is a issue about mysql on warming up but I haven’t found any concrete solution/advice to this.

The issue in my application is I am getting “504 Gateway Time-out” error because of this slow first query.

Any thoughts about myqsl warm up? and how to improve it.

Thanks

Rulix

I have a similar question as to what Chad asked about having multiple column keys. We need to build up queries based on columns a,b,c,d and in those queries the column ‘a’ will always be there.

So the queries can be like

WHERE a= AND b= AND c= AND d=
OR
WHERE a= AND c= AND d=
OR
WHERE a= AND b= AND d=

Would you recommend one composite index like KEY ‘my_key(‘a’,'b’,'c’,'d’) which can server my purpose efficiently?

Thanks in Advance
Manik

I have a table with over 80,000,000 rows containing GPS information latitude, longitude, date_time, vehicle_id and a few more columns.

I can’t seem to find a good index combination for a query bringing back around 7 days of results for a vehicle. Tried (date_time,vehicle_id) and MYSQL won’t use it, (date_time) and its bringing back thousands of rows under explain. Using innodb. Inserting up to 100 per second.

Query

select
g.date_time, g.speed,
g.odometer, gl.building_number_name,
gl.street, gl.town_city, gl.region, gl.postcode, io.high, i.id as is_idle, m.id as is_moving, o.id as over_speed,
l.name as location_name
from gps_positions g

inner join geocoded_locations gl on g.geocoded_location_id = gl.id
left join io_events io on g.id = io.gps_position_id
left join idling_events i on g.id = i.gps_position_id
left join moving_events m on g.id = m.gps_position_id
left join overspeed_events o on g.id = o.gps_position_id
left join locations l on l.id = gl.location_id
left join ios on io.io_id = ios.id

where
g.date_time between ’2009-10-15 00:00:00′ and ’2009-10-15 23:59:58′
and g.vehicle_id = 6261
and ( (io.high is not null and ios.gpio = or i.id is not null or m.id is not null or o.id is not null )
order by g.date_time;

Thanks, Dan

Indeed this is a hard choice. You may end up having a lot of indexes if you need to have a lot of different search combinations. Some of them (for example double ranges) may not be handled effectively by MySQL at all.

One of solutions we used when indexes can’t be constructed is using Sphinx (www.sphinxsearch.com) which can be thrown on the cluster easily and which can serve queries very fast even without having perfect indexes.