It’s an ancient tradition amongst MySQL DBAs to analyze slow query log files and sort the results by the so-called “index ratio” to find badly indexed queries. The “index ratio,” in this case, is a heuristic derived from the ratio of rows read to rows returned. The thinking goes that if a query is reading lots of rows and returning only a few, it might be missing an index.
The trouble with this is that it’s very prone to false positives: GROUP BY queries, for example, will often read lots of rows and aggregate them into few.
Instead of relying on a surrogate measure such as this sometimes-accidentally-right ratio, why not get the truth straight from the horse’s mouth? MySQL sets a flag in a query when it doesn’t use an index. And VividCortex sniffs this flag right off the wire and generates metrics from it.
To rank queries by their frequency of no-index-used flag, just go to Top Queries and change the Rank By menu to “Missing Indexes.” Voila! Bonus tip: explore the other options in the Rank By menu, too!
via Planet MySQL
December Tip of the Month: Find Queries that Lack Indexes