Supporting CDP components

The following table lists the components and the layout for a five node cluster.

Table 1. Cloudera Runtime component distribution for a five node Cloudera Observability On-Premises cluster
Service	Node 1 (All master components of all services)	Node 2, 3, 4 (Worker nodes + ZooKeeper + Cloudera Observability processing components)	Node 5 (Worker nodes + Cloudera Observability processing components + Cloudera Observability UI )
Cloudera Management	`Alert Publisher` `Event Server` `Host Monitor` `Reports Manager` `Service Monitor`
HBase	`Gateway` `Master` `Thrift Server (optional)`	`Gateway` `RegionServer`	`Gateway` `RegionServer`
HDFS	`Balancer` `Gateway` `NameNode` `NFS Gateway (optional)` `SecondaryNameNode`	`DataNode` `Gateway`	`DataNode` `Gateway`
Hive	`Gateway` `Metastore Server` `HiveServer`	`Gateway`	`Gateway`
Hue (Optional)	`Load Balancer` `Hue Server`
Impala	`Catalog Server` `StateStore`		`Impala Daemon`^*
Kafka		`Kafka Broker`
Phoenix	`Query Server`	`Query Server`	`Query Server` Configuring multiple Phoenix Query Server hosts reduces bottlenecks. The number of Phoenix Query Server hosts must be proportional to the number of Cloudera Observability On-Premises roles. For example, if you have roles on 5 nodes, at least 5 Query Servers are recommended for Phoenix. Cloudera Observability On-Premises internally balances loads on those hosts.
ZooKeeper		`Server`

^* To enable High Availability for Impala across daemons, you can manually configure multiple hosts by setting up HAProxy. For information, see Apache Impala documentation.