elasticsearch-watchdog/tmpl/watchdog.yml

A watchdog of elasticsearch - cluster nodes' statuses monitor, auto restart, keep PRIMARY node unique.

############# Elasticsearch Watchdog Configuration Example #############
# Overview of all various configuration settings.
#  YAML reference: http://en.wikipedia.org/wiki/YAML


############################## WATCHDOG ################################
# The name of current WATCHDOG, it is required.
#
watchdog.name: elasticsearch-server

# The frequency of watchdog patrolling.
#   https://www.npmjs.org/package/dynamic-timer
#
# it could be one of:
#   low       - lucas
#   medium    - fibonacci
#   high      - dayan
#   critical  - procession
#
# medium by default.
#
#watchdog.frequency: medium

################################ NODES #################################

# Both `nodes.elasticsearch` and `nodes.ssh` must be instances of Array and keep the same lengths.`
# And each host must have a ssh connection, e.g.:
#   nodes.elasticsearch: ["host1"]
#   nodes:
#     ssh:
#       - host: host1
#         ...
#

# Configure an list of nodes in the cluster of elasticsearch, e.g.:
#   nodes.elasticsearch: ["host1", "host2:port"]
#
# The port is optional, if it is not defined, 9200 will be used by default.
#
nodes.elasticsearch: ["host1", "host2:port2", "host3:port3"]

# Connect to the specific server by OpenSSH, it should be defined as:
#   nodes:
#     ssh:
#       - host:       192.168.1.100
#         port:       22
#         username:   root
#         password:   p@ssword
#       - host:       192.168.1.101
#         port:       22
#         username:   root
#         password:   p@ssword
#
# A plain text `password` in the configured file is extremely dangerous, but don't worry, try to encrypt
# your password through CLI like following:
#  $ watchdog pwd <PASSWORD>
# And copy the encrypted string to replace the `p@ssword`.
#
nodes:
  ssh:
    # defined an anchor.
    - &ssh1
      host:     host1
      port:     22
      username: root
      password: passw@rd
      es_stop:  kill -9 `cat /var/www/elasticsearch-1.3.2/pid`
      es_start: elasticsearch -d -p /var/www/elasticsearch-1.3.2/pid -Xmx8G -Xms8G
    # inherits the properties by anchor id - a YAML feature.
    - <<: *ssh1
      host:     host2
      password: c2162f2355d0430066194c8d356db578

    - <<: *ssh1
      host:     host3
      password: 75f0dc5bb33bbe48891f1175f69bf00c

################## ELASTICSEARCH CONDITIONS ######################

# All these conditions indicate whether node available or not, ElasticSearch nodes will be auto-restarted if unavailable.

# Server is running smoothly when the status of ElasticSearch matches this Setting,
# WATCHDOG will not auto-restart your ElasticSearch server.
# It could be "yellow", "green" or "red", and if not be defined, ["yellow", "green"] will be set by default.
#
# yellow  - A little buzy.
# green   - Smoothly.
# Red     - Very buzy, unresponsive sometimes.
#
#elasticsearch.status: ["yellow", "green"]

# The strategy of keeping primary node, it could be:
#   MS2M
#     the minority is subordinate to majority, if a node stands alone, WATCHDOG will try to shut it down, and restart it.
#   IP_ADDRESS
#    Always keep the primary node as a specific one, e.g.: "192.168.1.1".
#
# MS2M by default.
#
#elasticsearch.primary: MS2M

# The latency time of ElasticSearch stop and start.
# 5 seconds by default.
#
#elasticsearch.delay: 5s

# A value indicates whether auto-restart the ElasticSearch server or not, `true` by default.
# Auto-restart will be triggered in one of the following situation.
#   1. If ElasticSearch APIs, _cluster/health or _cluster/state can not be resolved(timeout).
#   2. Node status is not one of `elasticsearch.status`.
#   3. Master node does not conform to the `elasticsearch.primary`.
#
#elasticsearch.autorestart: true


############################## HTTP ###############################
# Http request options.
# All the time various such as `http.wait` and `http.timeout` support below formats:
#   h: hour, `1h` means 1 hour.
#   m: minute, `2m` means 2 minutes.
#   s: second, `3s` means 3 seconds.
#   {Number}: millisecond, `4000` means 4000 milliseconds(4 seconds).

# The latency time of Health check, just be used after ElasticSearch being restart.
# THE REBOOT TIME AND NODES' STATUSES WAS PERSISTED AS A JSON FILE, SO MAKE SURE YOUR PROJECT DIRECTORY IS WRITABLE.
# 20 minutes by default.
#
#http.wait: 20m

# The http `timeout`.
# 10 seconds by default.
#
#http.timeout: 10s

# The latency time when a http request error is caught.
# 5 seconds by default.
#
#http.delay: 5s

# The maximize number of times to connect to ElasticSearch server.
# 3 by default.
#
#http.retry: 3