[EXPORTER] Prometheus: Error on ingesting samples with different value but same timestamp

[Mifodix]

Describe your environment
opentelemetry-cpp 1.9.0, prometheus 2.43.0.

Steps to reproduce

	auto provider = std::make_shared<metrics_sdk::MeterProvider>();
	provider->AddMetricReader(std::move(prometheus_exporter));
        // ...
       	auto pluginCounter = metrics_api::Provider::GetMeterProvider()->GetMeter("test")->CreateUInt64Counter("plugin_counter3");
	while (1)
	{
		pluginCounter->Add(1);
		std::this_thread::sleep_for(std::chrono::milliseconds(500));
	}

PrometheusExporter generates such output:

# TYPE exposer_transferred_bytes_total counter
exposer_transferred_bytes_total 21388
# HELP exposer_scrapes_total Number of times metrics were scraped
# TYPE exposer_scrapes_total counter
exposer_scrapes_total 31
# HELP exposer_request_latencies Latencies of serving scrape requests, in microseconds
# TYPE exposer_request_latencies summary
exposer_request_latencies_count 31
exposer_request_latencies_sum 9972
exposer_request_latencies{quantile="0.5"} 276
exposer_request_latencies{quantile="0.9"} 524
exposer_request_latencies{quantile="0.99"} 553
# TYPE plugin_counter3 counter
plugin_counter3 155 1682692038761

Timestamp (1682692038761) for plugin_counter3 is not changing.

What is the expected behavior?
timestamp of plugin_counter3 is changed after each call Add() in cpp-code.

What is the actual behavior?
Timestamp is not changing. Errors in Prometheus logs:
msg="Error on ingesting samples with different value but same timestamp" num_dropped=1
Prometheus is ignoring changes of metric value.

Additional context
In my opinion fix must be in PrometheusExporterUtils::TranslateToPrometheus:

 * Helper function to convert OpenTelemetry metrics data collection
 * to Prometheus metrics data collection
 *
 * @param records a collection of metrics in OpenTelemetry
 * @return a collection of translated metrics that is acceptable by Prometheus
 */
std::vector<prometheus_client::MetricFamily> PrometheusExporterUtils::TranslateToPrometheus(
    const sdk::metrics::ResourceMetrics &data)
{

  // initialize output vector
  std::vector<prometheus_client::MetricFamily> output;

  for (const auto &instrumentation_info : data.scope_metric_data_)
  {
    for (const auto &metric_data : instrumentation_info.metric_data_)
    {
      auto origin_name = metric_data.instrument_descriptor.name_;
      auto sanitized   = SanitizeNames(origin_name);
      prometheus_client::MetricFamily metric_family;
      metric_family.name = sanitized;
      metric_family.help = metric_data.instrument_descriptor.description_;
      // ============= !!! FIX !!! ==============================================
      auto time = metric_data.end_ts.time_since_epoch();  // metric_data.start_ts.time_since_epoch();
      //======================================================================== 

[lalitb]

@Mifodix thanks for raising the issue. would you like to contribute the fix as you suggested ?

[renan028]

@lalitb
I had the same issue and the suggestion fixed the problem.
Thanks @Mifodix