Hortonworks installation

[#1]

I am going to install hortonworks cluster for my company. I was plan to install it with ambari, but somehow, I only got 1 namenode, 4 datanode, we dont even have secondary node.

I understood that ambari need a separate server to install agent on cluster hosts,

question is: Can I install ambari on the namenode, in other words, do I have to have separate node to install ambari?

please help.

thanks,

Robin

[#2 · Accepted solution]

@Robin Dong

Yes of-course you can install Ambari server on Namenode machine, basically if you want you can install everything on single node 🙂 Please check what is purpose of using hadoop, think about volume of input data, how much raw data you want to retain, how much output data you want to retain, what kind of processing you are going to do on Hadoop, what all ecosystems you are planning to use, by considering all these points decide hardware resources and distribute hadoop components amongst nodes in the cluster.

Hope this information helps!

[#3]

@Robin Dong

You can install the cluster using those nodes.

Don't install the components that you don't need as it can save you resources in the cluster for the components you need.

So...In you case

1 master node and 4 worker nodes. No edge node. This is old pic has it has Nagios and Ganglia which is replaced by ambari metrics server.

Node 1 - Ambari Server , Namode and Snode (or there is no harm having this on Node2 with DN),RM,Hive,Hive metastore

Node 2 to Node 4 - Datanode, Nodemanager and other processes.

Main question: Can I install ambari on the namenode, in other words, do I have to have separate node to install ambari?

Yes and you don't need separate server for ambari. The design for production environment is different from dev, qa and POC environments

[#4]

@Kuldeep Kulkarni and @Neeraj Sabharwal,

Hi Gurus, you two provided all info I am looking for. thank you so much for your help.

Seemed this community is a great place to get help and so many gurus out there...

Thanks again.

Robin

[#5]

@Robin Dong I am glad it helped. You can accept one the best answers to close the thread or we can close thread by assuming both answers are good. "Best practice"

[#6]

@Robin Dong

My pleasure 🙂

[#7]

yes, both answers provided me the best practice. I really appreciated.

we are having a meeting on Monday to decide which is the best ETL tool for load data from taradata database to hortonworks cluster, so any suggestions. Sparks or anything else.

I guess this is a new question, should I create another thread?

[#8]

Yes please... @Robin Dong