Vijay Samuel

@albertteoh @jpkrohling we are currently using spanmetrics processor to generate red metrics and ingest into prometheus. these metrics are in the scope of a service+operation context. so if we either...

> I wonder if prometheus federation at smaller scale maybe. but if we have several 100s of services, depending on what granularity the RED metrics are being generated by the...

@jpkrohling could you kindly help move this forwad? are there more approvals required if i may ask?

Bumping for interest again :)

Incorrectly raised in Grafana in the past: https://github.com/grafana/grafana/issues/29214

@bboreham thanks for getting back. I somehow missed this comment. The requirements that we intend to meet overall is that given a recording/alert rule, to add exemplars into the recorded...

@bboreham let's say there is a metric `latency` which has `pod`, `namespace`, `instance`, `le`, `http_path` and `http_verb` as labels. a recording rule `histogram_quantile(0.99, sum(rate(latency[5m])) by (le, http_path, http_verb, namespace)` would:...

@bboreham could you kindly take a look?

@bboreham thank you for the review. i have the same concern about how things like `label_replace` would work. I will do a rebase, add test cases and address all the...

> if I run a rule count by (job)({job!=""}) does it fetch and re-append every exemplar? unfortunately yes. i have contemplated on adding a sampling ratio (0-1) to only add...