To measure the number of errors in our system, we need to find the ratio of requests that had a successful response code to those that did not have a successful response code. We can find this with the following query sum(requests {code!="200"}) / (sum(requests {code="200"})) + sum(requests {code!="200"}).

This ratio is important to monitor. Computer systems fail and people make incorrect requests, but your ratio of 200 responses to non-200 responses should be relatively small.