Support Questions
Find answers, ask questions, and share your expertise

CDH 5 fresh installation


Hi folks,

I have a quick question as this is my first time to insall this MRv2 Cluster 

I do not have enough number of physical nodes, Can I install {Resource Manager RM node , Name Node} on same machine, another node for the secondary Name Node, and install the yarn-proxyserver, and mapreduce-historyserver on the same node of RM or this will kill the performance.


I am just following the installation steps here it looks like I need to have at least 4 nodes for just having one data node and the rest for the cluster management but I want to install {RM, NameNode, YARN-proxyserver, MapReduce-history server} on one machine and another machine to be dedicated for the secondary namenode. 





Small clusters usually don't have a lot of load, so it should be fine to put almost all of the singleton daemons (Cloudera Manager daemons, Resource Manager, NameNode, Job History Server, etc) on host 1, the SecondaryNameNode on host 2, and all slaves (DataNode, NodeManager, etc) on hosts 2-4. This is what the Cloudera Manager wizard will suggest to you automatically for a cluster of size 4.


With only 3 slave nodes, how much load can you really generate on the master roles like namenode and secondary namenode anyway? If you've got more than 20 then it can look like a concern. Again, when setting up your cluster initially, CM will recommend generally reasonable placement for your roles.