We are planning to mount Hadoop 2 in a cluster with the following machines:
- One machine for Management Server
- One machine for Namenode Server
- One machine for Resource Manager
- n machines for Datanodes
We have doubts where to run Flume agents in the cluster.
Is recommended to run collector tier Flume agents in the same machine, which one? Or is better to run one collector Flume agent on each Datanode?
Storage agents should run in Namenode machine?
Is correct or recommended in our cluster to configure collector tier agents sinks destination directly to HDFS storage (not using an Storage Tier)?
Thanks for help!