Comments on: 7 ways to convince MySQL to use the right index http://code.openark.org/blog/mysql/7-ways-to-convince-mysql-to-use-the-right-index Blog by Shlomi Noach Sat, 13 Mar 2010 09:46:07 +0000 http://wordpress.org/?v=2.9.2 hourly 1 By: Jim http://code.openark.org/blog/mysql/7-ways-to-convince-mysql-to-use-the-right-index/comment-page-1#comment-3222 Jim Thu, 13 Aug 2009 22:15:38 +0000 http://code.openark.org/blog/?p=695#comment-3222 This bug may be fixed in 5.1.37: http://dev.mysql.com/doc/refman/5.1/en/news-5-1-37.html We had EXACTLY the same problem as described in this blog, and upgrading to 5.1.32 seems to have fixed it. This bug may be fixed in 5.1.37: http://dev.mysql.com/doc/refman/5.1/en/news-5-1-37.html

We had EXACTLY the same problem as described in this blog, and upgrading to 5.1.32 seems to have fixed it.

]]> By: Artículos destacados de Abril 2009 | cambrico.net http://code.openark.org/blog/mysql/7-ways-to-convince-mysql-to-use-the-right-index/comment-page-1#comment-1819 Artículos destacados de Abril 2009 | cambrico.net Thu, 07 May 2009 07:48:44 +0000 http://code.openark.org/blog/?p=695#comment-1819 [...] formas de forzar el uso de un índice determinado, en openark.org. (en [...] [...] formas de forzar el uso de un índice determinado, en openark.org. (en [...]

]]> By: Interesting Items - April 27, 2009 | Oxidiser techblog http://code.openark.org/blog/mysql/7-ways-to-convince-mysql-to-use-the-right-index/comment-page-1#comment-1705 Interesting Items - April 27, 2009 | Oxidiser techblog Mon, 27 Apr 2009 10:07:11 +0000 http://code.openark.org/blog/?p=695#comment-1705 [...] 7 ways to convince MySQL to use the right index [...] [...] 7 ways to convince MySQL to use the right index [...]

]]> By: shlomi http://code.openark.org/blog/mysql/7-ways-to-convince-mysql-to-use-the-right-index/comment-page-1#comment-1588 shlomi Fri, 17 Apr 2009 02:50:16 +0000 http://code.openark.org/blog/?p=695#comment-1588 Hi nolan, Yes, I think this should work also. Since I don't have that DB available to me anymore, and since I cannot reproduce the problem, I can't test... If indeed it works (and if EJB supports it) - I like it better than any of the above solutions. Thanks Hi nolan,

Yes, I think this should work also. Since I don’t have that DB available to me anymore, and since I cannot reproduce the problem, I can’t test…
If indeed it works (and if EJB supports it) – I like it better than any of the above solutions.

Thanks

]]> By: nolan http://code.openark.org/blog/mysql/7-ways-to-convince-mysql-to-use-the-right-index/comment-page-1#comment-1586 nolan Thu, 16 Apr 2009 20:39:40 +0000 http://code.openark.org/blog/?p=695#comment-1586 What if you broke it into a query and subquery? i.e.: SELECT id from (SELECT id FROM data WHERE type=12345 AND level > 3) ORDER BY id This should make MySQL ignore the primary key in the subquery. I guess you'll also lose the primary key in the main query, but that just means sorting a list of ~100 entries, so it's not so bad. What if you broke it into a query and subquery? i.e.:

SELECT id from (SELECT id FROM data
WHERE type=12345 AND level > 3)
ORDER BY id

This should make MySQL ignore the primary key in the subquery. I guess you’ll also lose the primary key in the main query, but that just means sorting a list of ~100 entries, so it’s not so bad.

]]> By: Aaron Suggs http://code.openark.org/blog/mysql/7-ways-to-convince-mysql-to-use-the-right-index/comment-page-1#comment-1492 Aaron Suggs Mon, 06 Apr 2009 21:29:39 +0000 http://code.openark.org/blog/?p=695#comment-1492 Ah, you're right. Putting the primary key as the last column in a multicolumn index is redundant. I hadn't thought of that. Ah, you’re right. Putting the primary key as the last column in a multicolumn index is redundant. I hadn’t thought of that.

]]> By: shlomi http://code.openark.org/blog/mysql/7-ways-to-convince-mysql-to-use-the-right-index/comment-page-1#comment-1487 shlomi Mon, 06 Apr 2009 16:26:31 +0000 http://code.openark.org/blog/?p=695#comment-1487 Hi Aaron, Please see my previous comment: on InnoDB tables, and when the PRIMARY KEY is `id`, the `type` index is implicitly an index on (`type`, `id`). This is because of the implementation InnoDB does for secondary indexes, which point to PRIMARY key values. Hi Aaron,

Please see my previous comment: on InnoDB tables, and when the PRIMARY KEY is `id`, the `type` index is implicitly an index on (`type`, `id`). This is because of the implementation InnoDB does for secondary indexes, which point to PRIMARY key values.

]]> By: Aaron Suggs http://code.openark.org/blog/mysql/7-ways-to-convince-mysql-to-use-the-right-index/comment-page-1#comment-1485 Aaron Suggs Mon, 06 Apr 2009 12:34:44 +0000 http://code.openark.org/blog/?p=695#comment-1485 Another option is to get rid of the `type` index and add a multicolumn index on `type` and `id` (in that order, not `id` then `type`). With a multicolumn index, the plan for the original offending query would expect 110 rows, and wouldn't use filesort (since the index contains the sorted values). Another option is to get rid of the `type` index and add a multicolumn index on `type` and `id` (in that order, not `id` then `type`).

With a multicolumn index, the plan for the original offending query would expect 110 rows, and wouldn’t use filesort (since the index contains the sorted values).

]]> By: shlomi http://code.openark.org/blog/mysql/7-ways-to-convince-mysql-to-use-the-right-index/comment-page-1#comment-1479 shlomi Sun, 05 Apr 2009 17:38:11 +0000 http://code.openark.org/blog/?p=695#comment-1479 Hi Hannes, I did try it, but the query plan still went for the PRIMARY key only. It's interesting to note that, on InnoDB tables, the key on 'type' actually includes the 'id' column implicitly, since secondary indexes contains the value of the primary keys. If the optimizer chooses to use the 'type' index on "ORDER BY type, id", it can do so using index only (however, in our case we also had a filter on the 'level' column, which negates this optimization) Hi Hannes,

I did try it, but the query plan still went for the PRIMARY key only.
It’s interesting to note that, on InnoDB tables, the key on ‘type’ actually includes the ‘id’ column implicitly, since secondary indexes contains the value of the primary keys.
If the optimizer chooses to use the ‘type’ index on “ORDER BY type, id”, it can do so using index only (however, in our case we also had a filter on the ‘level’ column, which negates this optimization)

]]> By: Hannes http://code.openark.org/blog/mysql/7-ways-to-convince-mysql-to-use-the-right-index/comment-page-1#comment-1478 Hannes Sun, 05 Apr 2009 16:59:14 +0000 http://code.openark.org/blog/?p=695#comment-1478 My first guess would have been: ORDER BY type, id Have you tried this? My first guess would have been:

ORDER BY type, id

Have you tried this?

]]>