Ansible rolling restart of Elasticsearch Cluster

elasticsearch-handlers-main.yml

---
###
# Elasticsearch Rolling restart using Ansible
###

##
## Why is this needed?
##
#
# Even if you use a serial setting to limit the number of nodes processed at one
# time, Ansible will restart elasticsearch nodes and continue processing as soon as
# the elasticsearch service restart reports itself complete.  This pushes the cluster 
# into a red state due to muliple data nodes being restarted at once, and can cause
# performance problems.
#
##
## Solution:
##
#
# Perform a rolling restart of the elasticsearch nodes and wait for the cluster
# to stabilize before continuing processing nodes.
#
## 
## NOTES
##
#
# Our elasticsearch role and these handlers assume Ansible is running on one
# of the Elasticsearch cluster nodes (see the use of "localhost:9200" below)
#
# Variable definitions
#
# service - Our cluster uses three types of nodes:
#   'search' - Provides the front end Elasticsearch api to our applications
#   'ossec'  - Runs on our ossec masters to for input of ossec-hids logs and alerts
#   'data'   - Holds the cluster data
#
# 'search' and 'ossec' nodes are eligible to be masters.
#
# The elasticsearch role is called with the service variable set appropriately.
#
# Each node in our cluster is 'node.name'd {{ansible_hostname}}-{{service}} in its
# elasticsearch.yml configuration file.
#
# The handlers are chained together using notify to perform the following process:
#
# 1. Disable shard allocation on the cluster
# 2. Restart the elasticsearch node process
# 3. Wait for the node to rejoin the cluster
# 4. Enable shard allocation on the cluster
# 5. Wait for the cluster state to become green
#
# This keeps the cluster from entering a red state.

# Restart Elasticsearch node
# This handler is the entrance point for the chained set of handers
- name: restart elasticsearch-{{service}}
  debug: msg="Rolling restart of node {{ansible_hostname}}-{{service}} initiated"
  changed_when: true
  notify:
  - cluster routing none {{ansible_hostname}}-{{service}}

# Turn off shard allocation
- name: cluster routing none {{ansible_hostname}}-{{service}}
  local_action: "shell curl -XPUT localhost:9200/_cluster/settings -d '{\"transient\" : {\"cluster.routing.allocation.enable\" : \"none\" }}'"
  register: result
  until: result.stdout.find('"acknowledged"') != -1
  retries: 200
  delay: 3
  changed_when: result.stdout.find('"acknowledged":true') != -1
  notify:
  - do restart {{ansible_hostname}}-{{service}}

# restart elasticsearch process
- name: do restart {{ansible_hostname}}-{{service}}
  service: name=elasticsearch-{{service}} state=restarted
  notify:
  - wait node {{ansible_hostname}}-{{service}}

# Wait for Elasticsearch node to come back into cluster
- name: wait node {{ansible_hostname}}-{{service}}
  local_action: "shell curl -s -m 2 'localhost:9200/_cat/nodes?h=name' | tr -d ' ' | grep -E '^{{ansible_hostname}}-{{service}}$' "
  register: result
  until: result.rc == 0
  retries: 200
  delay: 3
  notify:
  - cluster routing all {{ansible_hostname}}-{{service}}

# Turn on shard allocation
- name: cluster routing all {{ansible_hostname}}-{{service}}
  local_action: "shell curl -s -m 2 -XPUT localhost:9200/_cluster/settings -d '{\"transient\" : {\"cluster.routing.allocation.enable\" : \"all\" }}'"
  register: result
  until: result.stdout.find("acknowledged") != -1
  retries: 200
  delay: 3
  changed_when: result.stdout.find('"acknowledged":true') != -1
  notify:
  - wait green {{ansible_hostname}}-{{service}}
  tags:
  - service

# Wait until cluster status is green
- name: wait green {{ansible_hostname}}-{{service}}
  local_action: shell curl -s -m 2 localhost:9200/_cat/health | cut -d ' ' -f 4
  register: result
  until: result.stdout.find("green") != -1
  retries: 200
  delay: 3

thanks!

hello!
i'm very curious about the context within you call this handler.
Do you use features like throttle serial ? This handler is in a role folder and is called from a role task ?
I can't reproduce the execution of this set of tasks node after node (i.e. every tasks are run on each node instead of the whole set of tasks is run on each node)

To be honest, I don't remember why I put this gist up. At the time, I was working heavily with elasticsearch and needed an ansible role that would reliably restart a cluster without downtime. I no longer work at the firm where I did this and don't have access to any of the other source code.

All right thank you for your answer!