lliknart / logstash-input-nats Goto Github PK

It looks like NATS#start will yield the started client to the given block, so we can use it directly with NATS::Client#subscribe instead of the global NATS::subscribe, which would enable multiple instances of this plugin to be able to run in the same Logstash process:

    NATS.start(@nats_config) do |nats_client|
      nats_client.on_error do |error|
        logger.error(error)
      end
      @subjects.each do |subject|
        logger.debug("Listening on [#{subject}]")
        nats_client.subscribe(subject, :queue => @queue_group ) do |msg, _, sub|
          @codec.decode(msg) do |event|
            decorate(event)
            event.set("nats_subject", sub)
            queue << event
          end
        end
      end
    end

Note: I also used the logger that is available on all Logstash plugins instead of Kernel#puts 😄

trapping TERM and INT in a Logstash plugin is considered bad-form, because these signals are used by Logstash for graceful shutdown. Binding them to a proc that calls Kernel#exit! is especially troublesome because that means inclusion of this plugin in one pipeline causes all pipelines to immediately and unsafely abort when receiving a signal that ordinarily would initiate a safe shutdown.

Hi,

Can this plugin be used with multiple servers? Like the nats-ruby client:

NATS.start(:servers => ['nats://127.0.0.1:4222', 'nats://127.0.0.1:4223']) do |nc|
  puts "NATS is connected to #{nc.connected_server}"

  nc.on_reconnect do
    puts "Reconnected to server at #{nc.connected_server}"
  end

  nc.on_disconnect do |reason|
    puts "Disconnected: #{reason}"
  end

  nc.on_close do
    puts "Connection to NATS closed"
  end
end

Hope to hear from you

changing the global error-handling of NATS to send Kernel#exit! means that any time NATS encounters an error, the entire Logstash process dies in an unsafe manner.