Expert Troubleshooting: Resolving MySQL Problems Quickly

Troubleshooting MySQL doesn't need to involve trial-and-error, and it doesn't need to take a long time. You need only two things: a good process and good tools. This session will show you both.
The process is paramount. Perhaps the most common thing I find when I'm called in to help with a difficult, long-standing problem is that the situation has been complicated significantly by trial and error. It's no longer possible to know what is the problem, and what's a symptom of changes that have been applied to the system. A good troubleshooter may develop theories and pursue intuition, but never without proof of a causal relationship, an informed prediction of the outcome, or ability to measure the results. This requires an understanding of the moving parts, and the flow of causes and effects through the system -- in other words, understanding of the system's architecture. It also requires the ability to measure and observe what the system is doing.
That's where the second pre-requisite comes in: good tools. Tools are necessary to change and measure systems. This session will introduce the best existing tools for helping to measure and observe systems, as well as new tools I have created when existing ones are not good enough. Examples include tools from Maatkit, a replacement for iostat (because it is not good enough), tools for measuring network behavior, and system summary tools.
How well does this work? At Percona, we've solved many longstanding, frustrating issues quickly with the tools and processes I'll explain. They aren't hard to use. It just takes a little practice, and you will be able to get to the bottom of problems very quickly. There will be case studies, but no hands-on lab work.