Implement abstraction of observability module with function returning MetricsEndpointProvider

[rgildein]

Context

The observability module should provide create_metrics_endpoint function, which will handle the following:

1. expose service port with ServicePort (using lightkube library)
2. patched exposed services with KubernetesServicePatch (using kubernetes_service_patch library)
3. creating MetricsEndpointProvider from provided list of pairs metrics_port and metrics_path (using prometheus_scrape library)
4. Return object created with step 3

Example:

def create_metrics_endpoint(charm, list of pairs of (metrics_port, metrics path), service_port) -> MetricsEndpointProvider:
    ...

What needs to get done

1. Add new component
2. Provide base structure for component
3. Add full unit tests coverage
4. [Optional] Add integration tests
5. Add README

Definition of Done

1. Component was tested, and it's working (either via integration tests or draft implementation in some charm)
2. There is proper README how to use the component

[syncronize-issues-to-jira[bot]]

Thank you for reporting us your feedback!

The internal ticket has been created: https://warthogs.atlassian.net/browse/KF-5710.

This message was autogenerated

[rgildein]

After internal discussion, we decide not to do it. The reasons are as follows:

• this abstraction will require to handle charm libraries, see #94
• some charms required to configure different types of ServicePort (e.g. mlflow), so either ServicePort will be provided as an input argument or function become to complex
• the use of KubernetesServicePatch in this abstraction represents the mixing of different logics and at the same time specific settings are sometimes required, e.g. refresh_event in mlflow
• without the two points above, this abstraction no longer makes sense, as its added value will be very small