eyeem.elasticsearch-restart

An Ansible role for performing safe Elasticsearch rolling restarts, without the cluster ever getting red.

As specified on the documentation page linked above, the current role performs the following steps:

• disable shard reallocation,
• restart the Elasticsearch service,
• wait for the machine to rejoin the cluster (checking number of nodes),
• reenable shard reallocation,
• wait for the cluster to return to green state.

Requirements

The role uses the shell module to perform curl requests to the local elasticsearch instance. You should ensure shell is installed on the host machines before invoking this role.

Role Variables

es_http_port: 9200
es_listening_host: localhost
es_service_name: elasticsearch

All commands will be executed locally on the host machine, calling curl on es_listening_host:es_http_port. The name of the elasticsearch service can be customized using the es_service_name variable.

Dependencies

The elasticsearch-restart role does not have any external dependencies.

Example Playbook

---
- hosts : "elasticsearch_cluster"
  serial: 1
  roles:
    - elasticsearch-restart

While you can simply use the elasticsearch-restart role like shown in the snippet above, you likely want to perform some other operations on each node before actually starting the controlled restart. To do so, you can use this role within a playbook that also takes care of pre- and post- tasks like registering and deregistering the node from monitoring and from any load balancers you might be using. Please refer to the Ansible documentation for further details.

---

- hosts : "elasticsearch_cluster"
  serial: 1
  pre_tasks:
    - include: monitoring_deregister.yml
    - include: balancer_deregister.yml
  roles:
    - elasticsearch-restart
  post_tasks:
    - include: balancer_register.yml
    - include: monitoring_register.yml

License

Apache v2

Contributors

• Michele Palmia (michele)
• Joé Schaul (joe)

Please email us @eyeem.com