I’ve heard from the guys at MySQL that the configuration of Heartbeat communication paths in ha.cf seems to be confusing to some. So, here’s our best practice summary:

• Never configure fewer than two communication paths. And I mean, never. Never. If you do, that’s an accident (read: split brain) waiting to happen.
• On the device that you run your DRBD replication over, if it uses a direct back-to-back connection to the peer node (like it should), configure a bcast link. Simple and easy to configure, and since you’re not sharing that link with anything other than DRBD replication, those broadcast packets won’t harm or …

Here is my notes on setting up replication on MySQL. In a lot of cases, that is not good enough, because it replicates EVERYTHING from the master to slave(s), whereas you may just want one or two databases replicated.

At first I thought I could just add this to /etc/my.cnf on the slave:
[mysqld]
replicate-do-db=MyDb

That didn’t work very well for statements like this, assuming you are in a database other than MyDb:

insert into MyDb.TableInMyDb values (SomeValue)

Fine, I thought, let me add this to slave’s my.cnf:

replicate-wild-do-table=MyDb.%

It’s not enough. For the Sql statement above, it still couldn’t catch and execute that statement.

After some searching and testing, I found what I was looking for. As far as I can tell, it works beautifully. All you need to do is to modify my.cnf on slave. Just add the databases to ignore in the list. Below is a …

Lenovo and Novell announce deal to preload SUSE Linux. Linspire CEO resigns. Red Hat teams with Creative Commons on Fedora content. (and more)

Lenovo and Novell To Offer Linux Preload on ThinkPad Notebooks, Novell / Lenovo (Press Release)

Fedora and Creative Commons Team Up To Deliver LiveContent Distribution, Red Hat (Press Release)

HP Drives Customer Adoption of Open Source and Linux in the Data Center, HP (Press Release)

LiMo Foundation Announces Membership Surge, LiMo Foundation, (Press Release)

…

There was a problem with improper characters apprearing in the content of one of the feeds. This caused a tag mismatch error in the XML parsing of the feed files and thus weren't showing up in Firefox and Thunderbird (possibly others). The feed has been removed and this has solved the problem.

I worked out these rules from speaking at MySQLcon, and attending OScon this year.

• Speaking at geek conferences is fun.
• Don't put your talk on your slides!
• Don't put everything on the slides, don't put the outline on your slides, don't put a lot of stuff on the slides.
• Put questions on the slides, instead. Each question is the one you answer while that slide is up.
• You can occasionally have an answer, a good quote, or a graphic.
• If someone can't QUICKLY write down what is on the slide, it's too complex.
• Do write speaker notes. Make them as complete as possible. Try to memorize them, but don't just recite them. You can refer to them if you have to during your talk.
• Practice your talk at least once. Check time. You may feel silly talking to yourself in front of the wall, but you will feel dumb instead if you run way …

Event: Zmanda Management Console for MySQL Backup - presented by Zmanda on Wednesday, August 22, 2007.

One of the simplest performance and load tests that can be performed on any web server is to measure the response time of a static HTTP GET request. If you measure how this response time degrades as more parallel connections issue the same request, you can get a basic understanding of the performance and scalability of a web server.

Anyone who's worked with Apache long enough is probably familiar with ab, the  load generator/benchmarking tool which is so wide-spread. This is a sad state of affairs as 'ab' is extremely flawed. I don't want to repeat it's flaws here, as Scott Oaks has done an excellent job of summarizing them. 

Another tool that is also popular is http_load, …

As a professional tester, I have a dream about a tool that will let me use my favorite SQL manager, and at the same time create tests ready to be plugged to the regression test suite.
I have been dreaming and planning about it, and finally the magic tool is shaping up.
I am mentoring a Summer of Code project, assigned to Charlie Cahoon. After a long preparatory work and much planning, the MySQL test creator is finally out!
Initially, it was planned as a separated tool, a replacement for the MySQL command line. The idea is that you write queries and eye the results, while the tool in background is …

Recently we were puzzled by question how query_cache works with column level privileges. The question was appeared as we discovered function query_cache_send_result_to_client is called before real parsing of query, so at the moment of execution the query_cache is not able to know which columns are accessed.
Looking into source code I found out that in the case when table has column level privileges - function query_cache_send_result_to_client just stopped and returns to usual execution. So initial assumtion was that query against table with defined column privilege is just not cached, though it is not reflected in documentation.
It was worth to test it experimentally and the reality turned out more interesting.

under root:

PLAIN TEXT SQL:

1. CREATE TABLE xp.test1 (id int, name varchar(23));
2. INSERT INTO xp.test1 VALUES (5, 'testit');
3. GRANT SELECT (id) ON TABLE …

If you read the comments on this blog, you’ll already know this…

Mark Leith, a Senior Support Engineer at MySQL, sent a link to a talk he gave (at a MySQL users conference) about MySQL for Oracle DBAs. It’s very informative, and I wish I had read it a long time ago. (I also wish I could have attended the talk!)

http://www.markleith.co.uk/dl/MySQL_for_Oracle_DBAs.pdf

He also points out that MySQL has prepared some training “boot camps” for Oracle DBAs about MySQL. I actually knew about this - I signed up right away! I’ll post about it after I take the class (Aug 16).

Thanks Mark!