tldr: not here (yet)
So you can already link between traces and logs, but what about connecting them to metrics? That means exemplars, individual data points with store extra labels, not just the aggregate data usually see in metrics.
A public design doc is available for prometheus.
The decision to record exemplars starts at the client,
so, API changes a necessary.
For a counter, it's add, but you store additional labels
(like traceID) for this particular point.
This will look something like:
1counter.AddWithExemplar(value float64, exemplar prometheus.Labels)
Next is to expose the data: exemplars are part of the OpenMetrics spec, only one exemplar per metric / bucket.
1foo_bucket{le=”0.1”} 8 <timestamp> # {id=abc} 0.043 <timestamp>
Since they're comments, they could be ignored,
but what good would that do?
Prometheus needs to store them, so we can query (and graph) them later.
Work is primarily in prometheus/prometheus#6635,
which didn't make it in to v2.25.0.
Final piece is grafana, which should have basic support in v7.4.0,
but it's kind of hard to test when storage doesn't work...