We wanted to share a document that describes some best practices, and solicit feedback for instrumenting workloads in cf-for-k8s. From a metric standpoint we are making good progress in cf-for-k8s on container metrics (see our metric proxy) and are starting to move towards metrics from other platform components. The major difference here is that we are no longer utilizing Loggregator envelopes, and are moving to a Prometheus exposition standard. We are here to help if your component is using something like statsd and needs to convert to Prometheus.
The document should be open to comments and suggestions so please make your comments there.