I haven’t really decided my schedule yet during the conference, but I thought I’d mention these sessions that look interesting to me.

I’m presenting a session on how to use Maatkit, which I think attendees will get a lot of benefit from.

• Tuesday
  • This is Not a Web App: The Evolution of a MySQL Deployment at Google (keynote). Mark Callaghan, need I say more?
  • MySQL and Search at Craigslist. Jeremy has gotten back into technical work after a career at Yahoo! that seemed to lead him away from his heart’s desire. I’d like to hear about the things he’s done, especially since I understand it involves replacing a lot of overloaded MySQL machines with a few underloaded Sphinx machines.
  …

I always say that the ultimate MySQL tuning script is an expert human. To that end, I generally try to build tools that help a human be more productive with the raw information from MySQL. One of the things we look at during a performance audit is the MySQL status counters. It’s useful to look at a) absolute values and b) several incremental snapshots. I’ve written a small shell script called “mext” that can make this a little easier.

It looks like this:

baron@kanga:~$ mext -- mysqladmin ext -ri1 -c3
Aborted_clients                               1      0      0
Aborted_connects                              0      0      0
Binlog_cache_disk_use                         0      0      0
Binlog_cache_use                              0      0      0
Bytes_received                             1167     35     35
Bytes_sent …

Some unsettling things happened in MySQL in the past week or so.

New storage engine not mentioned in the changelog

There’s a bit of a storm brewing over at the MySQL Performance Blog, where Vadim reports discovering a new storage engine added without mention in the 5.1.33 changelog. This is in defiance of the policy of not making changes in a production release. And it certainly belongs in the changelog — but there is no sign that anyone will remedy this problem.

Arjen Lentz, who is ex-MySQL and was Employee #25, reported a bug on the licensing of this storage engine. To my eyes, the engine’s license does not look right to include in a GPL database. Arjen agrees.

But the …

WARNING: nearly every word of this post is intentionally false. This is an April Fool’s joke.

There is so much misinformation out there about how MySQL works. If you’ve talked to someone from PostgreSQL, you’ve surely heard some of it: “MySQL doesn’t even have transactions,” for example. And this from a PostgreSQL user, who uses a database that doesn’t even run on Windows.

But even within the ranks of people who supposedly know MySQL’s workings, lots of people just get things dead wrong. I wanted to set the record straight here, so I thought I would give a little walk-through of how MySQL executes a query.

Let’s see it in action from start to finish.

SELECT SQL_NO_CACHE COUNT(col1) FROM users
WHERE userid IN(
   SELECT userid FROM othertbl
)
ORDER BY userid DESC;

The steps to execute this query are as …

SQL and Relational Theory

SQL and Relational Theory How to Write Accurate SQL Code by C. J. Date, O’Reilly 2009. Page count: 266 pages of “real” text, plus hefty appendixes. (Here’s a link to the publisher’s site: SQL and Relational Theory How to Write Accurate SQL Code).

This is a very important book for anyone involved with databases. Before I say why, I need to apologize to Mr. Date. I tech-reviewed part of the book and did not care for it. I am afraid I was quite a curmudgeon in my review comments. So, Mr. Date, if you’re reading this — I want to say I …

Here’s a quick productivity tip: when optimizing queries by rewriting them to different forms that should return the same results, you can verify that you get the same results by taking a checksum of them.

Just set your pager to md5sum:

mysql> pager md5sum -
PAGER set to 'md5sum -'
mysql> select * from test;
a09bc56ac9aa0cbcc659c3d566c2c7e4  -
4096 rows in set (0.00 sec)

What’s going on at your favorite open-source database developer mailing list? Non-scientific memory of what I’ve been seeing lately:

• The Drizzle list is talking about test coverage and completeness, and schedulers.
• The MySQL internals list has recently been dominated with talk about coding standards.
• The PostgreSQL hackers list is talking about a ton of stuff. I don’t even try to keep up.
• The Percona launchpad list is talking about bugs, xtrabackup, and more bugs and patches for them. There’s also a Percona Google Group …

Suppose you have a MySQL master-slave replication setup. It is very straightforward, nothing exotic at all. Imagine that it is the simplest possible setup. Everything is fine, there is nothing wrong or misconfigured with either server.

Now you add another slave, and replication on your existing slave fails. You did not change anything on your existing slave or master. How did this happen?

A while back I was at a seminar on migrating database applications to MySQL. A lot of the attendees were Oracle users. Based on their questions, comments and conversations during lunch, I made the following list of things Oracle users need to know about migrating to MySQL. Most of these are “gotchas” that would be contraindications or require some thought about a workaround.

Note: this is not meant to be MySQL-bashing. Some of these limitations are going to be fixed in future versions of MySQL, but they generally apply to current GA version 5.1. Some things are possible to achieve by choosing one particular way to use the server, at the exclusion of other things (e.g. Cluster rules out foreign keys, spatial data types rules out transactions). I am posting this list at the request of some Oracle DBAs who asked me to produce it for them.

1. Subqueries are poorly optimized.
2. Complex queries are a weak …

If you’ve been holding out for that golden opportunity, now’s a great time to get involved in Maatkit.

Until now I haven’t really made a conscious effort to open-source the decision process and get people involved; Maatkit has been largely driven by so-called “real-world needs,” as perceived through my little lens on the world (and emails from the whole Percona team telling me when something’s wrong). I guess I am likely to remain some kind of benevolent dictator, because I created Maatkit and historically I’m the main hacker. But it doesn’t have to stay that way, and the project and users will be better off if it doesn’t.

So I’ve been trying to break out of the rut of just having some little email exchange with people using the tools, and bring things onto the …