Dynamic Prometheus scraping across AKS pods using Fluent Bit Operator

[ptcbr]

We’re using Fluent Bit Operator on AKS and need to collect Prometheus metrics exposed by application pods across multiple namespaces.

The idea is to scrape /metrics endpoints from all pods that define Prometheus scrape annotations, similar to how Prometheus Service Discovery automatically detects and collects from annotated workloads.

The challenge is that the ClusterInput CRD only provides the prometheusScrapeMetrics input, which seems to require static host/port configuration. In our case, we need to scrape metrics from all workloads across the AKS cluster without manually defining each target.

Has anyone run into this limitation before?
Is there any existing Fluent Bit feature or pattern that can handle this kind of dynamic service discovery for Prometheus endpoints?