User Facing Metrics Service

[davidffrench]

User Facing Metrics Service

[davidffrench]

That is a fair assessment @emmanuelbernard .

I raised this functionality with the Observatorium team and shared this GitHub issue. The action is for the team to create a JIRA specifically for this which should elaborate on the job stories. Once that has been created, we should be able to close this specific issue.

[miguelsorianod]

Isn't this potentially covered by Observatorium??

[davidffrench]

Unfortunately, Observatorium does not currently authorise external users to their API. The idea is hopefully to have a conversation with the RHOBS team to extend Observatorium with this functionality.

[miguelsorianod]

Unfortunately, Observatorium does not currently authorise external users to their API. The idea is hopefully to have a conversation with the RHOBS team to build out this service.

I am not sure I understood this. Could you give a little bit more detail about it? At the moment we are able to write and read metrics to Observatorium.
Additionally, I understand you are talking about the Red Hat hosted Observatorium instance. Observatorium is an open-source project and its api mentions writing metrics(https://github.com/observatorium/api). I think it is important we differentiate the open-source software vs limitations that are imposed on Red Hat-hosted versions of the software

[davidffrench]

I think it is important we differentiate the open-source software vs limitations that are imposed on Red Hat-hosted versions of the software

Yes, you are correct. This GitHub issue might be better off as an internal issue. @emmanuelbernard Do you have any suggestions on how to track these types of issues?

[emmanuelbernard]

The way to track it is to think slightly generalized so that it addressed the need a user (Red Hat here) has.
A service will store lots of metrics and our de facto preference is the observatorium stack today (but any prometheus would do).
As a service you want to expose a user centric subset of these metrics and make sure they are exposed via an API with the proper authz. How do you do it assuming your metrics system does only expose metrics internally?

[emmanuelbernard]

One thing that is striking me is that this job story only refers to documentation. Wouldn't it be some endpoint / proxying / authz pre flight checking that a fleet manager implementor would rather have done for him or her?