Our application is encountering a high rate of errors that we consider irrelevant (they’re caused by a client-side bug in our system).
These irrelevant errors are distinguishable from relevant errors almost exclusively their error message. The HTTP code is no different from legitimate errors, and we care about all the other errors originating from requests against that same URL.
I haven’t been able to find any means of accomplishing this, so far.
The Golang agent can be configured to ignore certain HTTP codes. But as I say, HTTP is not a distinguishing feature of relevancy.
Some agents (Java and PHP?) support sophisticated server-side configuration (via the New Relic Web UI) that might help with this problem, but Golang isn’t one of them.
Also in the New Relic Web UI, errors can be “hidden”. But from what I can tell this doesn’t prevent the “hidden” errors from factoring into error rate calculations – it just removes the corresponding row from the list of errors in the UI of the error rate page. (I say this, because 99% of our errors are “hidden”, but the error-rate chart doesn’t move when toggling those errors “hidden”/“visible”.)
Can anyone advise?