Skip to content

Latest commit

 

History

History
394 lines (253 loc) · 26.6 KB

INSTALL_GCE.md

File metadata and controls

394 lines (253 loc) · 26.6 KB
Raw

ansible-hortonworks installation guide

  • These Ansible playbooks will build a Cloud environment on Google Compute Engine.
  • And then deploy a Hortonworks cluster (either Hortonworks Data Platform or Hortonworks DataFlow) using Ambari Blueprints.

Build setup

Before building anything, the build node / workstation from where Ansible will run should be prepared.

This node must be able to connect to the cluster nodes via SSH and to the Google Compute Engine APIs via HTTPS.

macOS

  1. Install the required packages

    brew install python
pip2 install virtualenv
pip2 install virtualenvwrapper

  2. Create and source the Python virtual environment

    virtualenv ~/ansible; source ~/ansible/bin/activate

  3. Install the required Python packages inside the virtualenv

    pip install setuptools --upgrade
pip install pip --upgrade   
pip install ansible apache-libcloud pycrypto

  4. Generate the SSH public/private key pair that will be loaded onto the cluster nodes (if none exists). Replace google-user with the non-root administrative user you want to use to login to the cluster nodes. You can also specify a different path for the key.

    ssh-keygen -q -t rsa -f ~/.ssh/id_rsa -C google-user

CentOS/RHEL 7

  1. Install the required packages

    sudo yum -y install epel-release || sudo yum -y install http://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpm
sudo yum -y install gcc gcc-c++ python-virtualenv python-pip python-devel libffi-devel openssl-devel sshpass git vim-enhanced

  2. Create and source the Python virtual environment

    virtualenv ~/ansible; source ~/ansible/bin/activate

  3. Install the required Python packages inside the virtualenv

    pip install setuptools --upgrade
pip install pip --upgrade   
pip install ansible apache-libcloud pycrypto

  4. Generate the SSH public/private key pair that will be loaded onto the cluster nodes (if none exists). Replace google-user with the non-root administrative user you want to use to login to the cluster nodes. You can also specify a different path for the key.

    ssh-keygen -q -t rsa -f ~/.ssh/id_rsa -C google-user

Ubuntu 14+

  1. Install required packages:

    sudo apt-get update
sudo apt-get -y install unzip python-virtualenv python-pip python-dev sshpass git libffi-dev libssl-dev vim

  2. Create and source the Python virtual environment

    virtualenv ~/ansible; source ~/ansible/bin/activate

  3. Install the required Python packages inside the virtualenv

    pip install setuptools --upgrade
pip install pip --upgrade
pip install ansible apache-libcloud pycrypto

  4. Generate the SSH public/private key pair that will be loaded onto the cluster nodes (if none exists). Replace google-user with the non-root administrative user you want to use to login to the cluster nodes. You can also specify a different path for the key.

    ssh-keygen -q -t rsa -f ~/.ssh/id_rsa -C google-user

Setup the Google account and credentials

To use the Ansible gce modules, you'll need to obtain the API credentials in the JSON format.

More details about how authentication to the Google Cloud Platform works is on the Ansible Guide.

  1. Create a Google Cloud Platform Service Account

    1. Go to the Service accounts page and login with your Google account.

    2. Decide on a project you want to use for the purpose of these scripts or create a new project in the All projects page.

    3. Once the project has been selected, click on the CREATE SERVICE ACCOUNT link.

    4. Give the Service account a name and a Role (recommended Role is Project -> Editor).

    5. Also select the Furnish a new private key option and JSON as the Key type. This will also initiate a download of the JSON file holding the service account's credentials. Save this file.

    6. If this is a new project, you'll also need to associate a Billing Account with the project (and create a new Billing Account if none exists). If this was done, confirm that everything works by going to the main Compute Engine page.

  2. Download JSON credentials

    If you haven't downloaded the JSON credentials file from the above step, or you already have a Service Account, go to the Credentials page and select Create credentials > Service account key.

    Select your Service Account and Create the JSON key.

  3. Upload the JSON credentials

    Once the JSON credentials file is obtained, uploaded it to the build node / workstation in any folder you prefer.

    This location will be referenced by an environment variable.

  4. Export the environment variables

    There are different ways to provide the credentials to the Ansible modules, each with its own advantages and disadvantages:

    • set variables directly inside the Ansible playbooks
    • populate a secrets.py file
    • setting environment variables

    All of these are explained in greater details on the Ansible Guide but for the purpose of this guide we'll use the following environment variables:

    • GCE_EMAIL: the email account associated with the project (can be found on the Service accounts page -> Service account ID column)
    • GCE_PROJECT: the id of the project (can be found on the All projects page)
    • GCE_CREDENTIALS_FILE_PATH: the local path to the JSON credentials file
    export [email protected]
export GCE_PROJECT=hadoop-123456
export GCE_CREDENTIALS_FILE_PATH=~/Hadoop-12345cb6789d.json

Upload the SSH public key to Google

Do the following to upload the SSH public key to the Google project.

This is based on Google's guide.

  1. Obtain the SSH public key

    Obtain the contents of the public key file (you can use the cat command).

    This can be an existing key or the one generated as part of the Build Setup, step 4:

    cat ~/.ssh/id_rsa.pub

  2. Add the key contents to Google Compute Engine

    Go to the METADATA PAGE and click on the SSH Keys tab.

    Click Edit and add the new key. When you paste the contents of the public key file obtained at the previous step, Google Compute Engine will automatically generate the Username, which is the non-root administrative user that is used to login to the cluster nodes.

    If you've used a different key than the one generated as part of the Build Setup, step 4, or you want to use a different user to login to the cluster nodes, replace the last bit of the key with the desired username.

    For the purpose of this installation guide, the username used is google-user.

Clone the repository

Upload the ansible-hortonworks repository to the build node / workstation, preferable under the home folder.

If the build node / workstation can directly download the repository, run the following:

cd && git clone https://github.com/hortonworks/ansible-hortonworks.git

If your GitHub SSH key is installed, you can use the SSH link:

cd && git clone [email protected]:hortonworks/ansible-hortonworks.git

Set the GCE variables

Modify the file at ~/ansible-hortonworks/inventory/gce/group_vars/all to set the GCE configuration.

cloud_config

This section contains variables that are cluster specific and are used by all nodes:

Variable Description
name_suffix A suffix that will be appended to the name of all nodes. Must be lowercase letters, numbers and hyphens or even the empty string ''.
region The Google Cloud Platform Region as described here.
zone The Google Cloud Platform Zone of the Region set above.
network The Google Cloud Platform Virtual Private Cloud network name as described here. It will be created if it doesn't exist.
subnet_name / subnet_cidr The Google Cloud Platform Subnet and IP range as described here. It will be created if it doesn't exist.
admin_username The Linux user with sudo permissions. This is the Username generated by GCE from the previously uploaded SSH public key.
ssh.privatekey Local path to the SSH private key that will be used to login into the nodes. This can be the key generated as part of the Build Setup, step 4.

firewall_rules

This is a separate section, just for firewall rules inside the network segment.

By default, the allow-internal type of rule should always be present, otherwise nodes would not be able to talk to each other.

Variable Description
name The name of the rule. The examples contain the network.name to be more easily identifiable.
allowed The protocol:ports to allow. The syntax allows multiple ports and protocols to be defined as long as they are split by ;. For example: tcp:80 or tcp:80,443 or tcp:80-800;udp:1-25.
src_range The IP source range, in CIDR notation. Any network / all sources can be identified with 0.0.0.0/0.

nodes config

This section contains variables that are node specific.

Nodes are separated by groups, each group defining a specific node role, for example master, slave, edge.

There can be any number of roles so other roles can be added to correspond with the required architecture.

And roles can have any names and any number of nodes but they should correspond with the host groups in the Ambari Blueprint.

Variable Description
role The name of the role. This will be appended to the cluster name in order to form a unique group in the same Google project. Other roles can be added to correspond with the required architecture.
count The number of nodes to be built with this role.
image The OS image to be used. More details here.
type The machine type / size of the node. A list of all the machine-types can be found here and the pricing here.
public_ip If the VM should have a Public IP assigned to it. Required if the build node does not have access to the private IP range of the cluster nodes.
root_disk By default, each Compute Engine instance has a single root Persistent disk that contains the operating system. The default size of the root volume / disk is 10GB, so use this variable to set the size of the root disk to the desired value. More details here about disk types and performance.
ambari_server Set it to true if the role also defines an Ambari Server. The number of nodes with this role should be 1. If there are more than 1 node, ambari-server will be installed on all of them, but only the first one (in alphabetical order) will be used by the Ambari Agents.

Set the cluster variables

cluster config file

Modify the file at ~/ansible-hortonworks/playbooks/group_vars/all to set the cluster configuration.

Variable Description
cluster_name The name of the cluster. This is also used by default in the cloud components that require uniqueness, such as the name of the nodes or tags.
ambari_version The Ambari version, in the full, 4-number form, for example: 2.6.2.2.
hdp_version The HDP version, in the full, 4-number form, for example: 2.6.5.0.
hdf_version The HDF version, in the full, 4-number form, for example: 3.1.2.0.
repo_base_url The base URL for the repositories. Change this to the local web server url if using a Local Repository. /HDP/<OS>/2.x/updates/<latest.version> (or /HDF/..) will be appended to this value to set it accordingly if there are additional URL paths.
java Can be set to embedded (default - downloaded by Ambari), openjdk or oraclejdk. If oraclejdk is selected, then the .x64.tar.gz package must be downloaded in advance from Oracle. Same with the JCE package. These files can be copied to all nodes in advanced or only to the Ansible Controller and Ansible will copy them. This behaviour is controlled by the oraclejdk_options.remote_files setting.
oraclejdk_options These options are only relevant if java is set to oraclejdk.
.base_folder This indicates the folder where the Java package should be unpacked to. The default of /usr/java is also used by the Oracle JDK rpm.
.tarball_location The location of the tarball file. This can be the location on the remote systems or on the Ansible controller, depending on the remote_files variable.
.jce_location The location of the JCE package zip file. This can be the location on the remote systems or on the Ansible controller, depending on the remote_files variable.
.remote_files If this variable is set to yes then the tarball and JCE files must already be present on the remote system. If set to no then the files will be copied by Ansible (from the Ansible controller to the remote systems).
external_dns This controls the type of DNS to be used. If yes it will use whatever DNS is currently set up. If no it will populate the /etc/hosts file with all cluster nodes.
disable_firewall This variable controls the local firewall service (iptables, firewalld, ufw). Sometimes, a local firewall service might run and block inter-node cluster communication. In these circumstances the local firewall service should be disabled as traffic rules should be provided by an external firewall such as Security Groups. Set to yes to disable the existing local firewall service if it blocks the inter-node cluster communication.
accept_gpl Set to yes to enable Ambari Server to download and install GPL Licensed packages as explained on the documentation.

security configuration

Variable Description
security This variable controls the Kerberos security configuration. If set to none, Kerberos will not be enabled. Otherwise the choice is between mit-kdc or active-directory.
security_options These options are only relevant if security is not none. All of the options here are used for an Ambari managed security configuration. No manual option is available at the moment.
.external_hostname The hostname/IP of the Kerberos server. This can be an existing Active Directory or MIT KDC. If left empty '' then the playbooks will install the MIT KDC on the Ambari node and prepare everything.
.realm The realm that will be used when creating service principals.
.admin_principal The Kerberos principal that has the permissions to create new users. No need to append the realm to this value. In case of Active Directory, this user must have Create, delete, and manage user accounts permissions over the OU container. If installing a new MIT KDC this user will be created by the playbook.
.admin_password The password for the above user.
.kdc_master_key The master password for the Kerberos database. Only used when installing a new MIT KDC (when security is mit-kdc and external_hostname is set to ''.
.ldap_url The URL to the Active Directory LDAPS interface. Only used when security is set to active-directory.
.container_dn The distinguished name (DN) of the container that will store the service principals. Only used when security is set to active-directory.
.http_authentication Set to yes to enable Kerberos HTTP authentication (SPNEGO) for most UIs.

ambari-server config file

Modify the file at ~/ansible-hortonworks/playbooks/group_vars/ambari-server to set the Ambari Server specific configuration.

Variable Description
ambari_admin_user The Ambari administrator's username, normally admin. This user and the password bellow are used to login to Ambari for API requests.
ambari_admin_password The Ambari password of the ambari_admin_user user previously set. If the username is admin and this password is different than the default admin, the ambari-config role will change the default password with the one set here.
ambari_admin_default_password The default password for the Ambari admin user. This is normally admin after Ambari is first installed. No need to change this unless there's a change in the Ambari codebase.
wait / wait_timeout Set this to true if you want the playbook to wait for the cluster to be successfully built after applying the blueprint. The timeout setting controls for how long (in seconds) should it wait for the cluster build.
default_password A default password for all required passwords which are not specified in the blueprint.
config_recommendation_strategy Configuration field which specifies the strategy of applying configuration recommendations to a cluster as explained in the documentation.
cluster_template_file The path to the cluster creation template file that will be used to build the cluster. It can be an absolute path or relative to the ambari-blueprint/templates folder. The default should be sufficient for cloud builds as it uses the cloud_config variables and Jinja2 Template to generate the file.

database configuration

Variable Description
database The type of database that should be used. A choice between embedded (Ambari default), postgres, mysql or mariadb.
database_options These options are only relevant for the non-embedded database.
.external_hostname The hostname/IP of the database server. This needs to be prepared as per the documentation. No need to load any schema, this will be done by Ansible, but the users and databases must be created in advance. If left empty '' then the playbooks will install the database server on the Ambari node and prepare everything with the settings defined bellow. To change any settings (like the version or repository path) modify the OS specific files under the playbooks/roles/database/vars/ folder.
.ambari_db_name,_username,_password The name of the database that Ambari should use and the username and password to connect to it. If database_options.external_hostname is defined, these values will be used to connect to the database, otherwise the Ansible playbook will create the database and the user.
.hive_db_name,_username,_password The name of the database that Hive should use and the username and password to connect to it. If database_options.external_hostname is defined, these values will be used to connect to the database, otherwise the Ansible playbook will create the database and the user.
.oozie_db_name,_username,_password The name of the database that Oozie should use and the username and password to connect to it. If database_options.external_hostname is defined, these values will be used to connect to the database, otherwise the Ansible playbook will create the database and the user.
.rangeradmin_db_name,_username,_password The name of the database that Ranger Admin should use and the username and password to connect to it. If database_options.external_hostname is defined, these values will be used to connect to the database, otherwise the Ansible playbook will create the database and the user.
.registry_db_name,_username,_password The name of the database that Schema Registry should use and the username and password to connect to it. If database_options.external_hostname is defined, these values will be used to connect to the database, otherwise the Ansible playbook will create the database and the user.
.streamline_db_name,_username,_password The name of the database that SAM should use and the username and password to connect to it. If database_options.external_hostname is defined, these values will be used to connect to the database, otherwise the Ansible playbook will create the database and the user.

ranger configuration

Variable Description
ranger_options These options are only relevant if RANGER_ADMIN is a component of the dynamic Blueprint stack.
.ranger_admin_password The password for the Ranger admin users (both admin and amb_ranger_admin).
.enable_plugins If set to yes the plugins for all of the available services will be enabled. With no Ranger would be installed but not functional.

blueprint configuration

Variable Description
blueprint_name The name of the blueprint as it will be stored in Ambari.
blueprint_file The path to the blueprint file that will be uploaded to Ambari. It can be an absolute path or relative to the roles/ambari-blueprint/templates folder. The blueprint file can also contain Jinja2 Template variables.
blueprint_dynamic Settings for the dynamic blueprint template - only used if blueprint_file is set to blueprint_dynamic.j2. The role names must match the roles from the inventory setting file ~/ansible-hortonworks/inventory/gce/group_vars/all. The chosen components are split into two lists: clients and services. The chosen Component layout must respect Ambari Blueprint restrictions - for example if a single NAMENODE is configured, there must also be a SECONDARY_NAMENODE component.

Build the Cloud environment

Run the script that will build the Cloud environment.

Set first the CLOUD_TO_USE environment variable to gce.

export CLOUD_TO_USE=gce
cd ~/ansible-hortonworks*/ && bash build_cloud.sh

You may need to load the environment variables if this is a new session:

source ~/ansible/bin/activate

Install the cluster

Run the script that will install the cluster using Blueprints while taking care of the necessary prerequisites.

Make sure you set the CLOUD_TO_USE environment variable to gce.

export CLOUD_TO_USE=gce
cd ~/ansible-hortonworks*/ && bash install_cluster.sh

You may need to load the environment variables if this is a new session:

source ~/ansible/bin/activate

This script will apply all the required playbooks in one run, but you can also apply the individual playbooks by running the following wrapper scripts:

  • Prepare the nodes: prepare_nodes.sh
  • Install Ambari: install_ambari.sh
  • Configure Ambari: configure_ambari.sh
  • Apply Blueprint: apply_blueprint.sh
  • Post Install: post_install.sh