We have installed cloudera manager on two different server and specified both IP address in each of agent file.
Will the agents commmunicate to both cloudera manager or not.
Note. - We have done this for to make cloudera manager HA
Please review the information we have provided in our documentation. We do not at present support active/active CM operation. You must setup Corosync and Pacemaker to allow for failover and resource control.
You don't need to install CM in two different nodes to enable HA.
I am just trying to explain in very very very high level
consider you have 5 nodes (before HA enabled)...
Node1: Master (Cloudera Manager installed), NameNode, Datanode
Node2: Master, Secondary namenode, Datanode
Node3: Slave, Datanode
Node4: Slave, Datanode
Node5: Slave, Datanode
1. HA will NOT have secondary name node. So enable HA in node2 instead of Secondary namenode
2. So you will have two namenodes (in both Node1 & Node2)
3. But always only one name node is Active and other one is Standby. You can confirm this in CM -> HDFS -> Instances (One is active and other is Standby)
4. So that if any point in time, if active breaks, the standby become active
Follow the link provided by @lhebert with above understanding.. Hope this will help you!!
In my case, Cloudera manager is on the same node where all other roles of CDH roles(Namenode,HbaseMaster, Spark,etc.) are installed.
My question is if this node get crashed how to I monitor and mange the cluster via cloudera manager.
Note - I have configured HA for namenode metadata and other roles. My concern is only for cloudera manager and it is using embedded postgresql.