Recently using vagrant to build a Hadoop cluster with 3 hosts, using Cloudera Manager to manage it, initially virtualized 4 hosts on my laptop, one of the most Cloudera manager servers, Several other running Cloudera Manager Agent, after the normal operation of the machine, found that the memory consumption is too strong, I intend to migrate two running Agent to another working computer, then use the Vagant package, and on another machine vagrant Box Add and vagrant up, after launch, login Cloudera Manager management interface to view, the results found that the two hosts migrated can not find, and constantly refresh the page, the display is the original running agent three machines, in addition to vm-cluster-node1 outside, Another host refresh in the Node2,node3 and node4 in turn, and not 4 machines at the same time, initially suspected because the package with Node1, so the agent has been included as Node1 agent configuration information.
So I ran the script to remove the agent service from both Node3 and Node4.
1 sudo /usr/sbin/service cloudera-scm-agent hard_stop_confirmed2sudo' cloudera-manager-* ' 3 sudo
Then I re-joined NODE3 and Node4 from the Cloudera Manager management interface, and before I joined, I removed the agent's machine from the new package, hoping to reuse it later.
Re-join the process is very long, you have to download more than 400 m manager agent package, can friends encounter the same problem? How can I not unload the package and only modify the configuration to make the migration successful??
Use vagrant to build a pit that the Hadoop cluster has stepped on