Allow Scraping of Node-Problem-Detector in kube-system-Namespace

[Christoph Andre]

How to categorize this issue?

/area monitoring
/kind enhancement

What would you like to be added:
Dear Gardener-Team,

we are using the CoreOS Kube-Prometheus-Stack to monitor our clusters, which brings a lot of exporters like node-exporter, node-problem-exporter etc.
However, when looking into the kube-system-Namespace we can see a lot of those monitoring components already present.
These are scraped by the Gardener-Prometheus and visualized by Plutono in the Gardener-Dashboard.
While it is possible, to pull most of the data out of your Prometheus, using federate, the node-problem-detector is missing.

In talks with your colleagues, it was discussed, that this component only reports to an operator, but not to prometheus.
Therefore there is also no service present.

However, for post-mortem analysis we are very much interested in this data and the currently only possible solution to get it, is deploying the node-problem-detector on our side again.

Is it possible to add a service resource into the shoot kube-system namespace so that the exporter can be reached from a customer-managed Prometheus?

Why is this needed:
Deploying these exporter twice will cost resources, which is imho uneccessary.