How to create exemplar and expose via PrometheusExporter?

[nemodev]

I am trying to make exemplars exported from my application's custom metrics, using PrometheusExporter.

I have an counter instance of io.opentelemetry.api.metrics.LongCounter type in my code and use add(Long l, Attributes attributes, Context context) method. I assume the Context parameter would include exemplar data for the metric, but I don't see exemplars from the exported metrics data.

I checked using below command.
curl -H 'Accept: application/openmetrics-text; version=1.0.0; charset=utf-8' http://localhost:6666/metrics

I can see my custom metric values there, but it doesn't have exemplar information like # {trace_id="...",span_id="..."}.

I enabled OpenMetrics option in the prometheus exporter config.

exporters:
  prometheus:
    endpoint: 0.0.0.0:6666
    send_timestamps: true
    enable_open_metrics: true
    resource_to_telemetry_conversion:
      enabled: true

Am I missing something or doing incorrectly?

I found an article about using client_java library from Prometheus project, but the example is using Prometheus metric interface (io.prometheus.client.Histogram), not opentelemetry metrics API. If possible, I don't want to introduce additional dependency(client_java) in my application and stick to opentelemetry metrics interface.

I also found another person(@irizzant) asked similar question in this channel and he got solved his issue, but I don't know how he generated exemplar in his code.

[mateuszrzeszutek]

Hey @nemodev ,

Currently the OTel Prometheus exporter only supports exemplars in histograms.
I'm not sure why we don't include exemplars when exporting other instruments' samples, I'll refer that question to @jack-berg

[jack-berg]

According to the spec, we're supposed to include exemplars on histograms and monotonic sums. We're currently not including them on monotonic sums (i.e. the aggregation of of counter instruments). I've opened this issue to track it.

[nemodev]

Hi @mateuszrzeszutek and @jack-berg ,

Thank you for the information. It is good to know that is current situation. I will wait for the counter metric support in the future.