safchain
commented
4 years ago Hello, I think both are valid approaches. For the second approach we can build a "logical NSM" topology add nodes and egdes with dedicated types so that we can filter them properly in the UI and with the API
In NSM we store traffic metrics in Prometheus - RX/TX bytes, packets and error packets. We have those for every pod-to-pod connection. We would like to show those in addition to the interface metrics. Or implement a Prometheus probe instead and build the network topology from the data stored in Prometheus. The latter doesn't look very native to the project to me but is still an option.
Any feedback and advice from the community are highly appreciated.