2

I want to create a Cloudera cluster using the quickstart VM image which you can directly download from cloudera´s web page (http://www.cloudera.com/downloads/quickstart_vms/5-8.html).

I have three virtual machines, I would like to have one master and two slaves. I´ve already configured them in order to have different hostnames and DNS connectivity between the 3 virtual machines. But when I try to add a new hosts, I usually get missing heartbeats from the other machines or when It doesn´t happen I get version mismatches and HDFS errors.

So, Is there any other configuration setting I should do before trying to add a new host?

daloman
  • 309
  • 3
  • 10

1 Answers1

0

If you want to do this using 3 quickstarts... don't. You're going to spend more time than create everything from scratch.

Nevertheless, the how-to:

  1. Cloudera quickstart remove any network hard written configuration to associate the FQDN to the real IP the first time. If your reboot your machine, depending how fast is starting the network, you will have your FQDN remapped to localhost. You have to deactivate this.
  2. Convert the default RPM installation to Parcel on all the nodes. There is an icon on the desktop. Why on all the nodes? because you're going to add your node using parcels and provoque conflicts between RPMs and Parcels. Another way would be remove all the RPMs package from Cloudera distrib, but with the risk to forget one of them...
  3. Remove cloudera-manager-server and cloudera-manager-agent from your slaves.
  4. Add your slaves to your master using the Add Host to cluster button using parcels.

And after having done all of this, you'll need to optimize Cloudera config to see everything green.

Personally, I prefer create a new template using Centos 7 rather Centos 6, implement 3 Vm fron the template, and install Cloudera from scratch.

Cheloute
  • 783
  • 2
  • 11
  • 27