Setup MapR Clusters with Puppet

Overview

This tool facilitates the setup of MapR clusters using Puppet. Starting from a fresh installed OS. It downloads & installs puppet and then runs 'puppet apply' (aka. puppet master-less mode) to do the setup. There is no need for any puppet infrastructure in place in advance.

The purpose is to provide a tool such that one can setup & configure an unsecure or MapR-SASL/Kerberos secured MapR cluster in the way he/she wants in a few steps. And once the configuration has been added into puppet hiera, the same/similar cluster can be setup with one command mc up.

This tool development is in progress. Not all ecosystem components have been added. You are welcome to contribute via pull requests :)

Steps

• Download or git clone this tool to a directory on your workstation
• Decide your cluster name and create the subdirectory: hieradata/<clustername>
• Review the variables in hieradata/default.yaml (e.g., cluster name, MapR core and MEP version, etc.) and override them in hieradata/<clustername>/cluster.yaml
  • You can set variables in hieradata/default.yaml directly if you manage only one cluster with this tool
• For each node in your cluster, create hieradata/<clustername>/<nodefqdn>.yaml, add the profiles/roles you want to have on that node
• For a secure cluster, generate the needed files and database on node70 (10.10.70.70):
  • Add your cluster nodes into /etc/hosts on node70
  • cd /root/intputfiles
  • For MapR SASL secured cluster, generate cldb key files:
    • /root/SetupMapRCluster/utils/create_cldb_keyfile.sh <clustername> <dnsdomain> node1 node2 ...
  • For Kerberos secured cluster, generate kerberos keytab files
    • cd /root/intputfiles
    • /root/SetupMapRCluster/utils/create_principal_keytab.sh <clustername> <dnsdomain> node1 node2 ...
  • To use the mysql server on node70 for hive metastore, run:
    • /root/SetupMapRCluster/utils/create_hive_meta.sh <clustername> <hivemeta_node>
  • Copy node70:/root/intputfiles/<clustername> to inputfiles/
• Copy the entire prepared directory to each individual node
• To install puppet and puppet modules, run under root or an account which can sudo to root: mc init
• To install configure and start the MapR cluster, run under root or an account which can sudo to root: mc setup
• Or combine both steps above with: mc up

FAQ

How to update properties in a hadoop xml configuration file?

Include the following into the corresponding module, for example for module profile::mapr::ecosystem::httpfs:

$file = "/opt/mapr/httpfs/httpfs-1.0/etc/hadoop/httpfs-site.xml"

package { ... }
->
profile::hadoop::xmlconf_property {
  # add a property
  "property_name1": file=>$file, value =>"value1", description=>"description1", ensure=>"present";
  # shorthand
  "property_name2": file=>$file, value =>"value2";
  # remove a property
  "property_name3": file=>$file, ensure=>"absent";
}

How to solve "shell-init: error retrieving current directory ... unhandled exception: boost::filesystem::current_path: No such file or directory"?

You may see the following error when run 'mc setup'/'mc up':

shell-init: error retrieving current directory: getcwd: cannot access parent directories: No such file or directory
2017-12-04 16:46:09.627479 FATAL puppetlabs.facter - unhandled exception: boost::filesystem::current_path: No such file or directory
terminate called after throwing an instance of 'boost::filesystem::filesystem_error'
what():  boost::filesystem::current_path: No such file or directory

The reason is that the working directory where you run mc has been removed in a different shell session or by a different app. Change to any existing directory should solve the issue.

Limitations

• This tool now supports only CentOS/RedHat.
• To setup a kerberized cluster, this tool uses node70 as the KDC. This works only in our UCS env. To do the setup in a cloud env, one can setup a simliar node like node70 with this tool, referring to hieradata/nodes/node70.ucslocal.yaml.

TODO

• Add more ecosystem components and their configurations