Not long ago we watched a production MySQL database melt down for sixteen minutes. The errors started as a trickle, a handful per minute, then fed on themselves: minute errors/min queries/s 0 5 15,000 <- a burst of work arrives 2 60 3,900 4 300 2,500 6 550 2,000 8 900 1,900 10 1,400 1,500 <- error peak = throughput trough 12 700 1,700 14 250 2,000 16 0 2,500 <- locks released, backlog drained 18 0

