Cluster configuration

Now we are ready to create and configure our first cluster with Jumbo!

Cluster creation and Jumbo context

First, lets enter the Jumbo shell and create our cluster:

[user@computer:~]$ jumbo
Jumbo v0.4.1
jumbo > create mycluster
Creating mycluster...
Cluster "mycluster" created (domain name = "mycluster.local").
jumbo (mycluster) >

After creating a cluster, the Jumbo context is set to this cluster. You can see the name of the cluster loaded (jumbo (mycluster) >). Use exit to reset the context, and then use to set the context to an existing cluster:

jumbo (mycluster) > exit
jumbo > use anothercluster
Loading anothercluster...
Cluster "anothercluster" loaded.
jumbo (anothercluster) > 

If you are in a hurry, use templates

When creating a cluster, you can reference a template. The cluster will then be configured in one second:

jumbo > create mycluster --template small-full
Creating mycluster...
Cluster "mycluster" created (domain name = "mycluster.local").
jumbo (mycluster) > listnodes
|   Name   |       Types        |      IP     | RAM (MB) | CPUs |
|  edge01  |        edge        | |   1024   |  1   |
| master01 | master, sidemaster | |   3072   |  1   |
| worker01 |       worker       | |   4096   |  2   |

If this is your first time using Jumbo, we recommend following this tutorial to learn everything you need to know, but you can jump straight to Cluster provisioning from here.
See the list of available templates on the Github repo (templates/docs).

Virtual machine creation

Now that we have created our cluster, lets add 3 virtual machines to it:

Adjust the RAM of VMs to your local machine!

jumbo (mycluster) > addnode master --types master --ip --ram 2048
Machine "master" added to cluster "mycluster".
jumbo (mycluster) > addnode smaster -t sidemaster -t edge
RAM (MB): 3072
Machine "smaster" added to cluster "mycluster".
jumbo (mycluster) > addnode worker -t worker --cpus 2
RAM (MB): 3072
Machine "worker" added to cluster "mycluster".

We now have all the nodes needed to deploy a functional Hadoop cluster. Use listnodes to see details about the nodes of the cluster:

jumbo (mycluster) > listnodes
|   Name  |      Types       |      IP     | RAM (MB) | CPUs |
|  master |      master      | |   2048   |  1   |
| smaster | sidemaster, edge | |   3072   |  1   |
|  worker |      worker      | |   3072   |  2   |

Service installation

A service can have dependencies to other services. A dependency is satisfied if the required service is installed and if the minimum required number for each component is installed. If the requirements to install a service are not met, Jumbo will tell you what services or components you have to install:

jumbo (mycluster) > addservice AMBARI
The requirements to add the service "AMBARI" are not met!
These services are missing:
jumbo (mycluster) > addservice ANSIBLE
Service "ANSIBLE" and related clients added to cluster "mycluster".
1 type of component auto-installed. Use "listcomponents -a" for details.

You can let Jumbo install all the dependencies with the tag --recursive:

jumbo (mycluster) > addservice AMBARI --recursive
The service AMBARI and its dependencies will be installed. Dependencies:

Do you want to continue? [y/N]: y
Service "AMBARI" and related clients added to cluster "mycluster".
Auto-installed the dependencies:
2 type of component auto-installed. Use "listcomponents -a" for details.

When installing a service, all its components are auto-installed on the best fitting hosts by default. You can avoid the auto-installation with the flag --no-auto.

Note that the service's clients will always be installed on all hosts (even with --no-auto) and on nodes created after the service installation. However you can use rmcomponent to delete them individually afterward.

A list of all the services supported by Jumbo is available here.

High Availability support
Some services support High Availability.
To install a service in HA, use the tag --ha with the command addservice.

Component installation

If you choose to not auto-install the components with the flag --no-auto, you have to manually add components with addcomponent on the machines of your choice. Use the command checkservice to see what components are missing for the service to be complete:

jumbo (mycluster) > checkservice HDFS
The service "HDFS" misses:
jumbo (mycluster) > addcomponent DATANODE -m worker
Component "DATANODE" added to machine "mycluster/worker".

Installation of all Hadoop services and components

We have to reproduce the same procedure of installation for the following services and components:

ServiceComponentsMachine type

Remove items

Use the commands rmnode, rmservice, or rmcomponent to remove items.

See what have been installed

Jumbo has list commands to describe the cluster state:

  • listclusters to list all the clusters and the services installed on each of them;
  • listnodes to list the VMs and their configurations;
  • listservices to list the status of each service installed on a cluster (complete or not);
  • listcomponents to list the components installed on a machine.