#! /usr/bin/env ruby # # es-node-graphite # # DESCRIPTION: # This check creates node metrics from the elasticsearch API # # OUTPUT: # metric data # # PLATFORMS: # Linux, Windows, BSD, Solaris, etc # # DEPENDENCIES: # gem: sensu-plugin # gem: rest-client # gem: json # # USAGE: # #YELLOW # # NOTES: # 2014/04 # Modifid by Vincent Janelle @randomfrequency http://github.com/vjanelle # Add more metrics, fix es 1.x URLs, translate graphite stats from # names directly # # 2012/12 - Modified by Zach Dunn @SillySophist http://github.com/zadunn # To add more metrics, and correct for new versins of ES. Tested on # ES Version 0.19.8 # # LICENSE: # Copyright 2013 Vincent Janelle # Copyright 2012 Sonian, Inc # Released under the same terms as Sensu (the MIT license); see LICENSE # for details. # require 'sensu-plugin/metric/cli' require 'rest-client' require 'json' # # ES Node Graphite Metrics # class ESNodeGraphiteMetrics < Sensu::Plugin::Metric::CLI::Graphite option :scheme, description: 'Metric naming scheme, text to prepend to queue_name.metric', short: '-s SCHEME', long: '--scheme SCHEME', default: "#{Socket.gethostname}.elasticsearch" option :server, description: 'Elasticsearch server host.', short: '-h HOST', long: '--host HOST', default: 'localhost' option :port, description: 'Elasticsearch port.', short: '-p PORT', long: '--port PORT', proc: proc(&:to_i), default: 9200 option :timeout, description: 'Request timeout to elasticsearch', short: '-t TIMEOUT', long: '--timeout TIMEOUT', proc: proc(&:to_i), default: 30 option :disable_jvm_stats, description: 'Disable JVM statistics', long: '--disable-jvm-stats', boolean: true, default: false option :disable_os_stats, description: 'Disable OS Stats', long: '--disable-os-stat', boolean: true, default: false option :disable_process_stats, description: 'Disable process statistics', long: '--disable-process-stats', boolean: true, default: false option :disable_thread_pool_stats, description: 'Disable thread-pool statistics', long: '--disable-thread-pool-stats', boolean: true, default: false def get_es_resource(resource) r = RestClient::Resource.new("http://#{config[:server]}:#{config[:port]}/#{resource}?pretty", timeout: config[:timeout]) JSON.parse(r.get) rescue Errno::ECONNREFUSED warning 'Connection refused' rescue RestClient::RequestTimeout warning 'Connection timed out' end def acquire_es_version info = get_es_resource('/') info['version']['number'] end def run # invert various stats depending on if some flags are set os_stat = !config[:disable_os_stats] process_stats = !config[:disable_process_stats] jvm_stats = !config[:disable_jvm_stats] tp_stats = !config[:disable_thread_pool_stats] es_version = Gem::Version.new(acquire_es_version) if es_version >= Gem::Version.new('1.0.0') stats_query_array = %w(indices http network transport thread_pool) stats_query_array.push('jvm') if jvm_stats == true stats_query_array.push('os') if os_stat == true stats_query_array.push('process') if process_stats == true stats_query_array.push('tp_stats') if tp_stats == true stats_query_string = stats_query_array.join(',') else stats_query_string = [ 'clear=true', 'indices=true', 'http=true', "jvm=#{jvm_stats}", 'network=true', "os=#{os_stat}", "process=#{process_stats}", "thread_pool=#{tp_stats}", 'transport=true', 'thread_pool=true' ].join('&') end if Gem::Version.new(acquire_es_version) >= Gem::Version.new('1.0.0') stats = get_es_resource("_nodes/_local/stats?#{stats_query_string}") else stats = get_es_resource("_cluster/nodes/_local/stats?#{stats_query_string}") end timestamp = Time.now.to_i node = stats['nodes'].values.first metrics = {} if os_stat metrics['os.load_average'] = node['os']['load_average'][0] metrics['os.load_average.1'] = node['os']['load_average'][0] metrics['os.load_average.5'] = node['os']['load_average'][1] metrics['os.load_average.15'] = node['os']['load_average'][2] metrics['os.mem.free_in_bytes'] = node['os']['mem']['free_in_bytes'] # ... Process uptime in millis? metrics['os.uptime'] = node['os']['uptime_in_millis'] end if process_stats metrics['process.mem.resident_in_bytes'] = node['process']['mem']['resident_in_bytes'] end if jvm_stats metrics['jvm.mem.heap_used_in_bytes'] = node['jvm']['mem']['heap_used_in_bytes'] metrics['jvm.mem.non_heap_used_in_bytes'] = node['jvm']['mem']['non_heap_used_in_bytes'] metrics['jvm.mem.max_heap_size_in_bytes'] = 0 node['jvm']['mem']['pools'].each do |k, v| metrics["jvm.mem.#{k.gsub(' ', '_')}.max_in_bytes"] = v['max_in_bytes'] metrics['jvm.mem.max_heap_size_in_bytes'] += v['max_in_bytes'] end # This makes absolutely no sense - not sure what it's trying to measure - @vjanelle # metrics['jvm.gc.collection_time_in_millis'] = node['jvm']['gc']['collection_time_in_millis'] + \ # node['jvm']['mem']['pools']['CMS Old Gen']['max_in_bytes'] node['jvm']['gc']['collectors'].each do |gc, gc_value| gc_value.each do |k, v| # this contains stupid things like '28ms' and '2s', and there's already # something that counts in millis, which makes more sense unless k.end_with? 'collection_time' metrics["jvm.gc.collectors.#{gc}.#{k}"] = v end end end metrics['jvm.threads.count'] = node['jvm']['threads']['count'] metrics['jvm.threads.peak_count'] = node['jvm']['threads']['peak_count'] end node['indices'].each do |type, index| index.each do |k, v| # #YELLOW unless k =~ /(_time|memory|size$)/ # rubocop:disable IfUnlessModifier metrics["indices.#{type}.#{k}"] = v end end end node['transport'].each do |k, v| # #YELLOW unless k =~ /(_size$)/ # rubocop:disable IfUnlessModifier metrics["transport.#{k}"] = v end end metrics['http.current_open'] = node['http']['current_open'] metrics['http.total_opened'] = node['http']['total_opened'] metrics['network.tcp.active_opens'] = node['network']['tcp']['active_opens'] metrics['network.tcp.passive_opens'] = node['network']['tcp']['passive_opens'] metrics['network.tcp.in_segs'] = node['network']['tcp']['in_segs'] metrics['network.tcp.out_segs'] = node['network']['tcp']['out_segs'] metrics['network.tcp.retrans_segs'] = node['network']['tcp']['retrans_segs'] metrics['network.tcp.attempt_fails'] = node['network']['tcp']['attempt_fails'] metrics['network.tcp.in_errs'] = node['network']['tcp']['in_errs'] metrics['network.tcp.out_rsts'] = node['network']['tcp']['out_rsts'] metrics['network.tcp.curr_estab'] = node['network']['tcp']['curr_estab'] metrics['network.tcp.estab_resets'] = node['network']['tcp']['estab_resets'] if tp_stats node['thread_pool'].each do |pool, stat| stat.each do |k, v| metrics["thread_pool.#{pool}.#{k}"] = v end end end metrics.each do |k, v| output([config[:scheme], k].join('.'), v, timestamp) end ok end end