David Ashpole
David Ashpole
A comment from the prometheus wg: We should consider splitting `build_info` out of `target_info`. There is a question about consistency across languages, and what the structure/overlap of those conventions are...
The two ideas from the spec sig today were: 1. Use the last word of the instrumentation scope name (and also, maybe allow overriding in the exporter w/ a mapping)...
I've updated the description and proposal to propose using the last word of the instrumentation scope.
> I have realized that I have no idea why this is a problem, why it may fail to create exporter or fail to send the metrics. Can you please...
I've updated the proposal to change the naming guidance for metrics instead of adding any conversion logic. I've also removed the `opentelemetry_scope_info` metric from this PR, and will propose that...
@tigrannajaryan It would be problematic if they had the same labels. E.g. ``` http_client_duration_sum{http_url="/foo"} 1 http_client_duration_sum{http_url="/bar"} 2 http_client_duration_sum{http_url="/bar"} 3 ``` Would be invalid. I also realized after the spec meeting...
> Only if they have the exact same timestamp, right? No. Prometheus/OpenMetrics metric points are required to have a [unique labelset](https://github.com/OpenObservability/OpenMetrics/blob/main/specification/OpenMetrics.md#metric) within the MetricFamily (i.e. metric name + label set...
> Does this solve the problem this PR tries to address or there is more to it? This solves it. Nothing more to it. > I wonder if there are...
> This is probably out of topic, bit I am not sure this is a good recommendation. The ability to aggregate http request time across libraries seems very useful to...
Added a changelog