Continuing the discussion from Ansible Role for Elasticsearch now available!:
I have many clusters running ES v1.x which were deployed by my own Ansible playbooks, which originated in the old Traackr version (https://github.com/Traackr/ansible-elasticsearch)
I have a new v2.2 cluster (testing for migration) and used the new 'official' role from https://github.com/elastic/ansible-elasticsearch
Everything is on Ubuntu 14.04 (trusty) at the moment.
During testing I was surprised to discover that the elasticsearch role java.yml
task installed openjdk-7-jre-headless
since the docs say to use Oracle JDK 8:
...as of this writing, it is recommended that you use the Oracle JDK version 1.8.0_73
https://www.elastic.co/guide/en/elasticsearch/reference/current/_installation.html
(That's in vars/Debian.yml
, the alternate vars/Redhat.yml
specifies java-1.8.0-openjdk.x86_64
)
I'm wondering what the motivation is in the Ansible for installing versions contrary to in-house recommendations.
Generally wondering, is this simply a question of what can be found in the Debian/Ubuntu repos?
But specifically wondering, should I modify that section to install Oracle 8 (I'm thinking yes).
I am concerned that there are Java args or environment settings hard-coded elsewhere in the Ansible role which presume OpenJDK...
Will dig, but wondering!
best regards,
aaron