Comments on: Detecting MySQL server problems automatically http://www.xaprb.com/blog/2012/09/06/detecting-mysql-server-problems-automatically/ Stay curious! Thu, 02 May 2013 12:36:53 +0000 hourly 1 http://wordpress.org/?v=3.5.1 By: Xaprb http://www.xaprb.com/blog/2012/09/06/detecting-mysql-server-problems-automatically/#comment-20216 Xaprb Fri, 07 Sep 2012 18:13:44 +0000 http://www.xaprb.com/blog/?p=2836#comment-20216 Little’s Law gives us a means to get average query response times over intervals of time using SHOW STATUS.

]]> By: Fernando ipar http://www.xaprb.com/blog/2012/09/06/detecting-mysql-server-problems-automatically/#comment-20215 Fernando ipar Fri, 07 Sep 2012 17:25:44 +0000 http://www.xaprb.com/blog/?p=2836#comment-20215 I think a big problem here is that statistical methods are usually about analyzing data to identify outliers (either from forecasts done from the same data, or from comparison against expected values).

For MySQL, in my experience, the most reliable indicator of trouble is query response time, and it’s impossible to get this out of MySQL directly.

So while we can apply the statistical methods to other indicators we can get out of MySQL, I can’t help to feel it would be like looking for your lost keys not where you dropped them, but where the light is better.

Now, if you want to do this without any client-side instrumentation (i.e new relic), I think the only hope right now would be a version of pt-query-digest that perpetually analyzes a slow log (probably a rate limited slow log) and calculating the response time for the 95th percentile of samples.

Finally, I second the request for making this available on github :)

]]> By: Jeremy Zawodny http://www.xaprb.com/blog/2012/09/06/detecting-mysql-server-problems-automatically/#comment-20209 Jeremy Zawodny Thu, 06 Sep 2012 23:17:22 +0000 http://www.xaprb.com/blog/?p=2836#comment-20209 I look forward to hearing what you’ve come up with.

Maybe put it on github? :-)

]]>