Anemometer
Collect and visulaize query performance & keep hist data
Raingage
Collect detailed forenisc data in response to an event
based on pt-stalk
Log with slow threashold=0.0 to log all queriers
pt-query-digent shows things in aggrate
pt-query-digest is run on each host and logged to a central server
Interactive dashboard for looking at data
HTTP GET based requests
Flot Graphs
git clone git://github.com/box/Anemeter.git anemometer
Sample 1% of production queries
Set log_slow_rate_limit
Run pt-query-digest in cron
In staging run a separate anemometer with slow_query=0.0 and leave on all the time.
pt-visual-explain tool
Finding phantom problems
Collect historical data
threads_running
Stalls show up a spikes in threads_ruiing
averave number of threads is <10