If you use Connect to move data from Kafka into InfluxDB (e.g. via this connector
), then you wouldn't need Telegraf since Connect would handle that step for you.
As for the choice of the stack, I am obviously biased, but Connect has a bunch of features that make it nice for this use case such as easy scalability, fault tolerance, delivery guarantees, pluggable serialization formats, and more listed in
http://docs.confluent.io/3.1.1/connect/intro.html Kafka also works really well for real time data integration, so you'll probably have more uses that Telegraf (which targets metrics) may not be a great fit for. Preferring Connect may reduce the number of technologies you need to learn & the number of systems you need to monitor.
-Ewen