daramao / apachebench_statsd_testing

Using ApacheBench / statsd to perform load tests.

Level 1

• Sign up for Datadog, get the agent reporting metrics from your local machine.
  Answer: I have set up the agent reporting metrics for my local machine: host:Daras-MacBook-Air-2.local

• Bonus question: what is the agent?
  Answer: The agent is a piece of software that collects events and metrics on the users hosts for monitoring and managing performance data.

• Submit an event via the API.
  Answer: To submit an event via the API, my code:

  require 'dogapi' api_key = 'DATADOG_API_KEY' dog = Dogapi::Client.new(api_key)

  response = dog.emit_event(Dogapi::Event.new('Level 1 question 3', :msg_title => 'Support Engineer Dara Mao'))

• Get an event to appear in your email inbox (the email address you signed up for the account with)
  Answer: By adding @[email protected] in the event body. My code:

  require 'dogapi' api_key = 'DATADOG_API_KEY' dog = Dogapi::Client.new(api_key)

  response = dog.emit_event(Dogapi::Event.new('Level 1 question 4 to @[email protected]', :msg_title => 'Support Engineer Dara Mao'))

Level 2

• Take a simple web app (in any of our supported languages) that you've already built and instrument your code with dogstatsd. This will create metrics.
  Answer: I have implemented dogstatsd for one of my rails project, Blackbox.

• While running a load test (see References) for a few minutes, visualize page views per second. Send us the link to this graph!
  Answer: This is the page view graph I generated using ApacheBench load test as follows:

  My code for page view:

  require 'statsd' statsd = Statsd.new statsd.increment('page.views')

  My code for load test:

  ab -n 157 -c 10 http://127.0.0.1:3000/periods/show ab -n 2000 -c 10 http://127.0.0.1:3000/periods/show ab -n 364 -c 100 http://127.0.0.1:3000/periods/show ab -n 10034 -c 1 http://127.0.0.1:3000/periods/show ab -n 100 -c 10 http://127.0.0.1:3000/periods/show ab -n 750 -c 100 http://127.0.0.1:3000/periods/show

• Create a histogram to see the latency; also give us the link to the graph. Bonus points for putting together more creative dashboards.
  Answer: This dashboard includes histogram along with page view load test.

My code for histogram:

 require 'statsd'
 statsd = Statsd.new
 begin = Time.now
 statsd.increment('page.views')
 end = Time.now
 frame = end - begin
 statsd.histogram('statsd.increment.latency', frame)

Level 3

• tag your metrics with support (one tag for all metrics)
• tag your metrics per page.
  Answer: I have added :tags =>['support, support:page1'] and :tags =>['support, support:page2'] to all the metrics for 2 pages of the app.
• visualize the latency by page on a graph (using stacked areas, with one color per page)
  Answer: Attached dashboard of graphs.

My code for load test:

 ab -n 120 -c 100 http://127.0.0.1:3000/periods
 ab -n 500 -c 10 http://127.0.0.1:3000/periods/show
 ab -n 45 -c 1 http://127.0.0.1:3000/periods
 ab -n 1532 -c 10 http://127.0.0.1:3000/periods/show
 ab -n 3495 -c 1 http://127.0.0.1:3000/periods
 ab -n 1200 -c 100 http://127.0.0.1:3000/periods/show

Level 4

• count the overall number of page views using dogstatsd counters.
• count the number of page views, split by page (hint: use tags)
• visualize the results on a graph
• Bonus question: do you know why the graphs are very spiky?
  Answer: Attached dashboard of graphs. The spikes in the graphs are due to time lapse in page load.

Level 5

• Write an agent check that samples a random value. Call this new metric: test.support.random
  Answer: My code for random_test.py:

  import random import statsd

  from checks import AgentCheck class RandomCheck(AgentCheck): def check(self, instance): self.gauge('test.support.random', random.random()) self.event({ "api_key": "DATADOG_API_KEY", "msg_title": "testing random AgentCheck", "tags": "test.support.random" })

        print(random.random())
  
      from statsd import statsd
      statsd.gauge('test.support.random', random.random())
  

  My code for random_test.yaml:

  init_config: default_timeout: 5

  instances: - url: http://127.0.0.1:3000/periods

• Visualize this new metric on Datadog, send us the link.
  Answer: I ran a few random tests, see attached dashboard of graphs.