Closed Raynos closed 9 years ago
Looks good in general. I think airlock would give us a way to do better errors vs warns because then we can error on thresholds and warn on individual failures.
@sh1mmer
I think generally what we want is a way to turn all warnings into stats and set thresholds on spikes in something like graphite & nagios or equivelant tooling.
@Raynos agreed, I think the point is distinguishing expected failures, repeated expected failures that are bad because they breached a threshold vs. failures which should never happen (errors)
cc @kriskowal @sh1mmer