.p50 metric is filtered out and does not make it to our observability tooling, renaming it to .median instead
the metrics seem to overflow, it's a bit unclear but a first attempt at a fix is to ignore CreationTimestamp is the zero value (it's unclear when this would actually happen, looking at kubectl get rr -A -o yaml that doesn't seem to be the case, maybe an unmarshalling issue between the k8s api server and our clients?).
After this PR
==COMMIT_MSG==
Rename p50 to median, attempt to address duration overflow
==COMMIT_MSG==
Before this PR
.p50
metric is filtered out and does not make it to our observability tooling, renaming it to.median
insteadCreationTimestamp
is the zero value (it's unclear when this would actually happen, looking atkubectl get rr -A -o yaml
that doesn't seem to be the case, maybe an unmarshalling issue between the k8s api server and our clients?).After this PR
==COMMIT_MSG== Rename p50 to median, attempt to address duration overflow ==COMMIT_MSG==
Possible downsides?