Closed Harwayne closed 3 years ago
Doesn't this correspond to filtering out 2xx responses and summing the values in a group by clause containing all the labels apart from response_code/response_code_class?
Created a similar issue in knative/eventing based on Adam's suggestion that it could be worthwhile exposing these metrics in OSS: https://github.com/knative/eventing/issues/4831
Doesn't this correspond to filtering out 2xx responses and summing the values in a group by clause containing all the labels apart from response_code/response_code_class?
I don't think so. The only events that would affect reply-event-count
are those that already are in the 2xx status code of event-count
, because the only way a reply is even attempted is if the request to the subscriber succeeds and it replies with an event.
Essentially, the flow of the event is something like:
event-count
and dispatch-latency
metrics.reply-event-count
metric, suggested in this issue (name is just an example, not a real suggestion).This issue is stale because it has been open for 90 days with no
activity. It will automatically close after 30 more days of
inactivity. Reopen the issue with /reopen
. Mark the issue as
fresh by adding the comment /remove-lifecycle stale
.
Problem We have metrics for Triggers sending events to the subscriber,
event-count
anddispatch-latency
. But we don't have any metrics around sending replies back into the Broker. After #2107, if the Trigger's reply request does not succeed, the overall delivery is considered a failure and the event is sent to the subscriber again. We need to have metrics so that the user understands what went wrong.Exit Criteria Triggers expose metrics around sending replies to the Broker's Ingress.
Additional context (optional)
event-count
is all that is really required, butdispatch-latency
may also be useful too.