How to setup a flume + kafka for a real time and streaming

[#1]

we plan to setup a flume + kafka cluster for our realtime and spark streaming system.

Questions here is that Kafka and flume are already in the cluster as a service when install hortonworks,

1. We like to have kafka server on a separate machine, not on any data node in the cluster, so how to move kafka to new machine. can I remove/add kafka on new machine from ambari?

2. We need few Kafka brokers (more machines add to this service) and all of these machines need to be configured as client node, right?

3. Is possible to add all brokers to ambari/kafka?

Any idea on how to do this?

[#2]

How have you installed HDP/HDF?

When you do a standard Ambari install of HDP it will ask you where you want to put each service, and you can choose to put Kafka on separate nodes from the other services.

You can also move Services using the Ambari interface, generally by putting them in maintenance mode on a particular node, stopping them, then selecting to move them to another node.

I would suggest that if you need more Kafka brokers in the same environment you should have them all controlled by the same Ambari, yes.