Jack Berg

Do you have a concrete example of a type of instrumentation that would benefit from this type of aggregation? > where every observation has a unique attribute set, with a...

I think this is a great topic to discuss. This problem isn't unique to the collector. Java has: - A general purpose [jmx metric gatherer](https://github.com/open-telemetry/opentelemetry-java-contrib/tree/main/jmx-metrics). - A general purpose [micrometer...

> I'm wondering how much of a problem this actually is. Even if some attributes change occasionally, target_info still are time series with constant values of 1, which should be...

> I think it is still a problem from timeseries continuity perspective. If I am looking at the CPU usage timeseries for a particular host I don't expect the timeseries...

> Also, what about Prometheus remote-write exporter, which is push, not pull? The [SDK prometheus exporter](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk_exporters/prometheus.md) "MUST NOT support [Push mode](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#push-metric-exporter)".

>1. Limiting data loss when round-tripping between Prometheus and OpenTelemetry. On this note the round trip behavior of `sdk w/ prometheus exporter -> collector w/ prometheus receiver -> otlp` is...

> This is not true tbh, and _info metrics are common to have joins on. Good to know! > we can always use metric_relabel_rules to drop the labels that have...

An initial implementation of cardinality defenses was [recently implemented](https://github.com/open-telemetry/opentelemetry-java/pull/3831) in java. Im summary, each view has a max number of metric streams. Currently, this max is hardcoded to be 2000,...

The java SDK changed its behavior after I posted that "we aggressively forget metric streams - any metric stream that doesn't have recordings during a collection is forgotten". We do...

@marcalff see [this comment](https://github.com/open-telemetry/opentelemetry-specification/issues/2729#issuecomment-1222584006). The current thinking is that any boolean environment variable should be named `*_ENABLED_*` or `*_DISABLED_*` such that if a default value of false aligns with the...