Hawkular - Monitoring Rails App using Hawkular Metrics

Setting up Hawkular Metrics

There are several ways of setting up Hawkular Metrics standalone server.

Build & deploy - follow Hawkular Metrics Readme
Docker image - start Hawkular Metrics Docker Image
Public instance - use http://livingontheedge.hawkular.org/hawklar/metrics which runs full Hawkular including Metrics. But be sure your data will get lost at some point as this instance gets rebuilt/cleaned up from time to time.

Monitoring Rails Application

There are several ways to monitor your Rails application’s performance. The way being shown here is pretty easy to implement, but requires modifications in Rails application. It takes advantage of around_filter of main ApplicationController. Within this filter we’re going to measure each controller’s response time to handle request.

Rails Demo Application

I’ve created simple Rails4 application called rails-demo.

I added the dependency on hawkular-client gem.

The Application controller code is pretty easy

require "hawkularclient"

class ApplicationController < ActionController::Base

  # Prevent CSRF attacks by raising an exception.
  # For APIs, you may want to use :null_session instead.
  protect_from_forgery with: :exception


  class MetricsReporter

    def initialize
      tenant_id = "hawkular.org"
      # create hawkular metrics client (1)
      @client = Hawkular::Metrics::Client::new("http://localhost:8080/hawkular/metrics","","",{:tenant => tenant_id})
      # create tenant in case it does not exist
      if @client.tenants.query.select { |tenant|  tenant.id == tenant_id}.empty?
      	@client.tenants.create(tenant_id)
      end
    end

    def around(controller, &block)
      # measure measure measure
      now = Time.now
      block.call
      time_spent = Time.now - now
      # metric ID must be unique within tenant (2)
      gauge = "App.#{controller.class.name}:#{controller.params[:action]}"
      # make sure metric definition is created (3)
      @client.gauges.create({:id => gauge, :tags => {:app => "App"}}) rescue
      # report time in milliseconds
      @client.gauges.push_data(gauge ,{:value => time_spent * 1000})
    end

  end

  # enable filter
  around_filter MetricsReporter::new

end

We create Hawkular Metrics @client. Point it to your instance URL, leave user/pass empty (as this has not been implemented yet) and set tenant option.
We encode our application Name, controller name and controller action into metric ID. So we have different metrics for listing objects and ie. editing an object.
We need to create metric definition with our tags, so we’re able to find all metrics related to rails-demo app later. This line is not required in case you don’t want to assign tags to your metric definitions.

Let’s start Rails dev environment

gem install rails
git clone https://github.com/lzoubek/rails-demo.git
cd rails-demo
bundler install
bin/rake db:migrate RAILS_ENV=development
rails server

By now, WEBrick should be listening on localhost, open http://localhost:3000/products and in case there is no error displayed congratulations, we are monitoring rails-demo! Let’s start hitting the UI with some more requests using helper bash script located in rails-demo repository

./hit.sh

At this time we are simulating traffic on rails-demo application and pushing controller’s response times to Hawkular Metrics service.

The Results

Currently Hawkular Metrics (standalone) has no UI. But there is a bunch of useful angular components like hawkular-charts, so I wrote simple UI which can at least find metric definitions by tags and draw metric data into line chart.

To watch metric data from rails-demo application open Hawkular Metrics UI Demo, fill in Metrics API URL (must be reachable from browser), Tenant ID and tags.

What’s next?

As you may have already noticed, Hawkular Metrics server perfomance can have negative impact on rails-demo application performance because we synchronously push the data within around_filter code. Even the fact, that rails application must be modified in order to be monitored might not be a good thing. In the next blog post I’ll write a simple service that is going to parse rails log files and report data to Hawkular Metrics. In fact, it would be awesome to reuse scout plugins and forward their reports into Hawkular Metrics.

Having metrics is only one piece in puzzle. Integration with Hawkular Alerts will let you detect performance regressions or unexpected loads.

Monitoring Rails App using Hawkular Metrics

Setting up Hawkular Metrics

Hawkular Ruby Client

Monitoring Rails Application

Rails Demo Application

The Results

What’s next?