my understanding of the ideal reference architecture for hadoop is 1 management node with 3 data nodes, with all servers connected via their own switch for a private network, with multiple ports bonded for throughput.
data and tools still need to get in and out so one of the ports on the management node should be attached to the public intranet.
data nodes will send results through the management node as well as receive data from it, so they need no public intranet access (aside from IMM ports).
is this a correct understanding?
This topic has been locked.
2 replies Latest Post - 2013-01-23T17:31:04Z by SystemAdmin
Pinned topic do data nodes need public IPs?
Answered question This question has been answered.
Unanswered question This question has not been answered yet.
Updated on 2013-01-23T17:31:04Z at 2013-01-23T17:31:04Z by SystemAdmin
Re: do data nodes need public IPs?2013-01-22T23:36:39Z in response to SystemAdminHi Mike,
Correct, data nodes should communicate over a private network.
Please see the diagram at this link: