In some cases, users want to know the count as it is exported by other meter registries (e.g. OTLP).

The count exported by other registries depends on what is expected with that registry. Some use a cumulative count, some use a delta count, some report each increment, some might use a step-normalized throughput. Since LoggingMeterRegistry is a StepMeterRegistry its meters are step meters, which for count means delta counts. But if you're trying to compare it with PrometheusMeterRegistry, for example, that is going to have cumulative counts instead. If we made it clear the logged count is a step count (delta count), that might help alleviate confusion, but it would still be hard to meaningfully compare it with other registries that aren't also a normal StepMeterRegistry. Ultimately, I'm not sure using the LoggingMeterRegistry as a means to tell what is exported by other registries is the best approach. Some registries themselves have logging that can be enabled to log the payload sent to the backend, if that's the level of debugging desired. I guess it might be helpful to understand more the intended use case here for the LoggingMeterRegistry. That said, I'm not opposed to showing a count, but we would want to take the above into consideration and not encourage people to use the LoggingMeterRegistry for some purpose it isn't fit for.

Ultimately, I'm not sure using the LoggingMeterRegistry as a means to tell what is exported by other registries is the best approach. Some registries themselves have logging that can be enabled to log the payload sent to the backend, if that's the level of debugging desired.

In our use case, we want to use:

• OtlpMeterRegistry for PROD environments that sends metrics to OpenTelemetry collector (itself sending metrics to Dynatrace).
  OtlpMeterRegistry is configured to produce delta metrics as required by Dynatrace.
• LoggingMeterRegistry for DEV environments and local workstations where Dynatrace is not used.

We considered logging at the level of OpenTelemetry collector for DEV environments but we wanted to avoid extra network costs between applications and collector.
Moreover, metrics logged in collector aggregate metrics from all applications in cluster and it makes it very complex for developers to find the metric they are interested in.

We don't plan to use LoggingMeterRegistry to compare metrics with other registries, I agree that it is better to have logging capabilities in each registry (if needed).
Our goal is only to have similar numbers in both approaches.

If we made it clear the logged count is a step count (delta count), that might help alleviate confusion

I also agree that using delta_count=1 instead of just count=1 will make it more clear that the count displayed in logs is not cumulative.

Sounds reasonable to me. Would you like to make a pull request for it?

Sounds reasonable to me. Would you like to make a pull request for it?

Yes I will.

Thanks