This would be really nice for observability of internal traffic.
Some open questions here:
- are the netflow boxes and also the Analytics pipelines involved going to be okay if we are sending a great number of more flows?
- can we get hostnames and/or cluster names rather than just IP addresses?
- are there any special worries around addressing of k8s pods? (we might need to map them from a pod IP --> a service name at flow creation time?)