Hi there!

We manage our clusters as node pools split up between interactive workloads and batch processing workloads, and as a result it's valuable for us to be able to aggregate things like resource usage grouped by node pool so we can understand things like resource contention at the level of a node pool.

Today it's possible to embed labels from a node if you are scraping a target discovered through node service discovery (with __meta_kubernetes_node_label_<labelname>, but node labels aren't available to you if you discover a workload through another SD method (in the case of an agent daemonset this would most commonly be pod).

It would be great if the grafana agent could expose node labels when running in host_filter mode, either through something like __meta_kubernetes_node_label_<labelname>, or even through a special set of environment variables like META_NODE_LABEL_<LABELNAME>.

Thanks!