class Smith::Riemann

Constants

SERVICE_NAME

Public Class Methods

new() click to toggle source
# File bin/riemann-smith, line 23
def initialize
  @riemann = Riemann::Client.new(:host => options[:host], :port => options[:port])
end

Public Instance Methods

alert(service, state, description, extra={}) click to toggle source

Construct an alert messages

# File bin/riemann-smith, line 99
def alert(service, state, description, extra={})
  {:tags => options[:tags], :ttl => options[:ttl], :service => service(service), :state => state(state), :description => description}.merge(extra).tap { |a| logger.info { "#{a[:service]}: #{state(state)}" } }
end
check(response) click to toggle source

Checks that the list of agents return from the list command matches the list of know agents and composes an appropriate alert.

# File bin/riemann-smith, line 70
def check(response)
  options[:group].map do |group|
    begin
      group_agents = options[:agents] + agent_group(group)
      running_agents = parse_response(response)

      group_agents.map do |agent_name|
        if running_agents.has_key?(agent_name) && running_agents[agent_name].any? { |a| a[:state] == "running" && running?(a[:pid]) }
          alert(agent_name, :running, "Agent running")
        else
          alert(agent_name, :critical, "Agent not running")
        end
      end
    rescue RuntimeError => e
      alert("Group #{group}: #{e.message}", :critical, e.message)
    end
  end.flatten
end
on_check(&blk) click to toggle source
# File bin/riemann-smith, line 64
def on_check(&blk)
  @on_check = blk
end
options() click to toggle source
# File bin/riemann-smith, line 117
def options
  @options ||= begin
    OptionParser.accept(Pathname) {|p,| Pathname.new(p) if p}

    defaults = {:interval => 30, :ttl => nil, :timeout => 11, :agents => [], :group => [], :tags => [], :host => 'localhost', :port => 5555, :agency_timeout => 60}
    defaults.tap do |options|
      parser = OptionParser.new do |opts|
        opts.separator "\n"
        opts.set_summary_indent "    "

        opts.banner = "\nUsage: #{opts.program_name} OPTIONS"
        opts.on_head "\n  Periodically lists the running agents and sends the result to riemann."

        opts.on("--ttl <i>", Integer, "TTL for events") { |t| options[:ttl] = t }
        opts.on("--interval <i>", Integer, "Polling interval in seconds (default #{options[:interval]})") { |t| options[:interval] = t }
        opts.on("--timeout <i>", Integer, "Agency timeout (default #{options[:timeout]})") { |t| options[:interval] = t }
        opts.on("--host <s>", String, "Riemann host (default #{options[:host]})") { |v| options[:host] = v }
        opts.on("--port <i>", Integer, "Riemann port (default #{options[:port]})") { |v| options[:port] = v }
        opts.on("--tags <tag1,tag1,...>", Array, "Tags to add to the alert") { |t| options[:tags] = t }

        opts.on("--agents <agent1,agent2,...>", Array, "Agents to monitor") { |v| options[:agents] + v }
        opts.on("--agent-group  <group name>>", Array, "The name of an 'agent group' to monitor") { |g| options[:group] += g }

        opts.separator "\n"
      end

      parser.parse!
    end
  end
end
parse_response(response) click to toggle source

Returns a Nested hash representing the running agents.

# File bin/riemann-smith, line 90
def parse_response(response)
  split_response = response.split(/\n/).map(&:strip)
  ((/^total/.match(split_response.first)) ? split_response[1..-1] : []).each_with_object(Hash.new { |h,k| h[k] = []}) do |e, acc|
    fields = e.split(/\s +/)
    acc[fields.last] << { :state => fields[0], :uuid => fields[1], :pid => fields[2], :description => fields.last }
  end
end
run() click to toggle source
# File bin/riemann-smith, line 27
def run
  Smith::Messaging::Sender.new(Smith::QueueDefinitions::Agency_control.call) do |queue|
    on_check do |check|
      @riemann << check
    end

    queue.on_error do |channel, channel_close|
      logger.fatal { "Channel error: #{channel_close.reply_code}: #{channel_close.reply_text}." }
      Smith.stop
    end

    EM.add_periodic_timer(options[:interval]) { run_check(queue) }
  end
end
run_check(queue) click to toggle source
# File bin/riemann-smith, line 42
def run_check(queue)
  @on_check.call(alert(:heartbeat, :running, "riemann-smith heatbeat", :ttl => options[:interval] * 2))

  queue.on_timeout(options[:timeout]) do |message_id|
    @on_check.call(alert(:agency, :timeout, "Agency timeout."))
  end

  queue.on_reply(:auto_ack => true, :reply_queue_name => "riemann.#{Digest::SHA1.hexdigest($$.to_s)}") do |reply_payload, r|
    check(reply_payload.response).each do |agent|
      @on_check.call(agent)
    end
  end

  queue.consumer_count do |count|
    if count > 0
      queue.publish(Smith::ACL::AgencyCommand.new(:command => 'list', :args => ['-a', '-l']))
    else
      @on_check.call(alert(:agency, :critical, "Agency not running."))
    end
  end
end
running?(pid) click to toggle source

Return true if the given pid exists in the process table.

# File bin/riemann-smith, line 104
def running?(pid)
  !(pid.nil? || Sys::ProcTable.ps(pid.to_i).nil?)
end
service(name) click to toggle source

Return the service name

# File bin/riemann-smith, line 109
def service(name)
  "#{SERVICE_NAME} #{name}"
end
state(state) click to toggle source
# File bin/riemann-smith, line 113
def state(state)
  {:timeout => "timeout", :running => "ok", :critical => "critical"}[state]
end