We need to collect and store metrics on Cassandra so that we can report on system health, detect problems, and determine when we to scale out. I think we could implement a custom metrics reporter. See this article for some background. I am referring specifically to DropWizard's metric reporter which is pluggable and is used extensively by Cassandra. Our metrics reported would be packaged with Cassandra and run in-process with minimal overhead. It would use the REST API to report data.
For OpenShift and ManageIQ deployments, we could include the customer report with Cassandra that it is automatically deployed. For other deployments, it may require some additional, manual steps.