grakn

Installs and configures GRAKN.AI on Ubuntu servers.

Requirements

• Ansible >2.3.1
• Tested on Ubuntu 16.04

The role needs to be run as root or with 'become: yes'.

Grakn itself requires:

• Oracle Java 8 on the target server.
• Cassandra
• Redis
• Zookeeper
• Kafka

For Analytics, you will need:

• Apache Spark cluster
• Apache Hadoop node

Role variables

For defaults values of the variables below, see defaults/main.yml.

Grakn version:

grakn_version: "0.14.0"

Backend configuration:

storage_hostname: 10.0.0.1,10.0.0.2,10.0.0.3
storage_cassandra_replication_strategy_class: org.apache.cassandra.locator.SimpleStrategy
storage_cassandra_replication_strategy_options: "replication_factor,1"

See Cassandra documentation for more information.

Auxiliary server configuration:

kafka_hosts: ['localhost']
zookeeper_hosts: ['localhost']
redis_host: "localhost" # single server

Java options passed directly to the JVM on the command line:

grakn_extra_options: "-Xmx12g -XX:+HeapDumpOnOutOfMemoryError"

Log level for Grakn:

grakn_log_level: "INFO"

Deploying custom releases

Your package needs to be named grakn-dist-SOME_VERSION.tar.gz.
This can be obtain by compiling Grakn from source

Download from URL

By default, the role will download the release available on GitHub and install it on the server.
To download from a custom location, add this to group_vars/all:

grakn_version: SOME_VERSION
grakn_download_url: "https://example.com/path/to/grakn-dist-SOME_VERSION.tar.gz"

You must specify the version so the playbook extracts the correct directory from archive and deploys the matching configuration.

Upload from local machine

The distribution archive must be located in files/ of the playbook's directory.
Add the following to your group_vars/all:

upload_release: true
grakn_version: SOME_VERSION

If you want to specify an alternative directory, use:

grakn_package_directory: "/full/path/to/tar/"

Upgrade / re-deploy Grakn

We currently don't have a way of checking if the installed and uploaded version of Grakn match for idempotence so to force a re-deploy, add this variable to your Grakn hosts:

redeploy: true

From the CLI you can run it as follows:

ansible-playbook -i inventory grakn_test.yml --tags role:grakn -e "{redeploy: true}"

Example playbook:

playbook.yml

- name: Grakn deployment
  hosts:
    - "grakn"
  become: yes
  roles:
    - { role: grakn }

group_vars/all

storage_hostname: "10.0.0.1,10.0.0.2,10.0.0.3"
zookeeper_hosts: [ 10.0.0.11 ]
kafka_hosts: [ 10.0.0.21,10.0.0.22,10.0.0.23 ]
redis_host: "10.0.0.31"

License

Apache 2.0