Migrating from prometheus client - MultiProcessCollector feature

[pavolloffay]

Is your feature request related to a problem?

Hello, we are considering migrating metrics instrumentation from the prometheus client. However the project depends in MultiProcessCollector:

• https://prometheus.github.io/client_python/multiprocess/
• https://github.com/prometheus/client_python/blob/master/prometheus_client/multiprocess.py#L19C7-L19C28

The feature does not seem to be available in the OpenTelemetry python.

Describe the solution you'd like

Support multi process mode in the OTEL python SDK for metrics

Describe alternatives you've considered

No response

Additional Context

Slack

Would you like to implement a fix?

None

[aabmass]

Support multi process mode in the OTEL python SDK for metrics

It sounds like you want to continue using Prometheus exporter/being scraped, is that right?

If sending OTLP to the prometheus server is an option for you (push), you can add a unique identifier to the service.instance.id to identify each process.

[pavolloffay]

The use case we solve with the Multiprocess Mode is to avoid reporting duplicated metrics by each process. The registry reports a single metric from multiple processes.

[aabmass]

If you're using the Prometheus exporter, it is not supported today #3747

What I mentioned above is your best option today. There are several issues in the backlog already for improving this with push metrics e.g. #3307.

[pavolloffay]

Thanks for the confirmation. We required prometheus exporter.

If sending OTLP to the prometheus server is an option for you (push), you can add a unique identifier to the service.instance.id to identify each process.

In case of OTLP exporter, can the SDK aggregate metrics from multiple processes like the prometheus client multiprocess/ feature?