Purpose

Gather statistics from a running Sauce Labs VM. Intended for Sauce Labs support only and offered with no guarantees of any kind. Customers: Don't use this. It might use all your minutes, exceed your concurrency, fail your tests and be mean to your dog, and you'll have bought it upon yourself.

(Also don't write to support asking about this they'll give you blank looks and ask me about it and I'll say it's unsupported and they'll close your ticket and you'll be sad. I don't want you to be sad.)

Description

Navi consists of two parts; the server and the script.

Server

The server is run on a known address and serves the script. It also listens for results and displays them in its log.

The server is all ready to run on Heroku; This is often the easiest way to get it publicaly routable. You don't even need a sane Ruby runtime to push the script to Heroku.

Script

The script is a python script that runs as a prerun executable. Because it needs to run for the duration of the test, it has to run in the background. Because it runs in the background, it can't write information to the test logs (ask Miketwo for why if you'd like some shouty despair).

The script runs a terminal command at a regular interval. The results of that command are then POSTed back to the webserver along with a timestamp.

The script configured currently will ping a domain 5 times every 5 seconds. You'll need to pass the domain as the first argument to the prerun executable. Load ping_script.py

The script can also check active, free and in use memory for Mac every second; load memory_script.py

Requirements

To run on Heroku

A free Heroku account
The Heroku CMD installed
Being logged into the Heroku CMD with heroku login

To run on another server

Ruby 2.x installed
Bundler installed (with gem install bundler)
The host IP or URL known
Possibly some SSL stuffs? IDK I didn't try this way.

Operating

Establishing the server

Add an app on the Heroku dashboard
Checkout this repo
CD into the repo directory
heroku git:remote -a your_app_name
git push heroku master
The app should be running on https://your_app_name.herokuapp.com

If it deployed successfully, your git output will include:

remote:        https://your_app_name.herokuapp.com/ deployed to Heroku
remote:
remote: Verifying deploy... done.

Check the application logs at https://dashboard.heroku.com/apps/your_app_name/logs and you should see output like this:

2017-03-31T01:14:48.245418+00:00 heroku[web.1]: State changed from down to starting
2017-03-31T01:14:50.249406+00:00 heroku[web.1]: Starting process with command `bundle exec thin start -R config.ru -e production -p 8398`
2017-03-31T01:14:54.563892+00:00 heroku[web.1]: State changed from starting to up

Setting up the Sauce Labs test

Set the prerun desired capability, ensuring it runs in the background:

caps['prerun'] = {executable: 'https://your_app_name.heroku.com/memory_script', background: true} (Example for Ruby)

Run It

Open the log viewer for your app at https://dashboard.heroku.com/apps/your_app_name/logs
Run your Sauce Labs test

Logs

Heroku's webapp allows you to see application logs at https://dashboard.heroku.com/apps/your_app_name/logs.

Verify your app fetched the prerun script

2017-03-31T01:17:27.599052+00:00 app[web.1]: 162.222.73.110 - - [31/Mar/2017:01:17:27 +0000] "GET /memory_script.py HTTP/1.1" 200 1814 0.0064

Verify your app is sending something for logging

2017-03-31T01:17:30.042449+00:00 app[web.1]: 162.222.73.110 - - [31/Mar/2017:01:17:30 +0000] "POST /stat/memory HTTP/1.1" 200 - 0.0008

Actual log data

2017-03-31T01:17:30.041904+00:00 app[web.1]: Thu Mar 30 18:17:22 2017 - active:656
2017-03-31T01:17:30.041930+00:00 app[web.1]: Thu Mar 30 18:17:22 2017 - free:255
2017-03-31T01:17:30.041931+00:00 app[web.1]: Thu Mar 30 18:17:22 2017 - total:1341.06640625

Customizing the script

The python script is served from logstats.py and is currently hardwired to only send memory statistics. This is because the webserver only knows how to nicely display memory stats.

To log different data, you will need to:

Create a new method that executes whatever commandline command you need data from, returning it either as a dictionary of interesting values or a single text blob
Change send_memory_to_server to call your new method
Change create_post_params to properly encode your data
Add a new route in routing.rb to accept your new data
Change the URL posted too on line 11 of logstats.py to match the new URL

Future plans

Other statistic scripts

I can immediately think of how to gather MTR, traceroute and ping data during tests.

We could also track CPU usage & open files.

logstats.py should be an ERB file so the servername is populated automagically.

Parameterized scripts

Allow the parameters of the prerun URL to re-configure the supplied script, eg:

https://your_app_name/script?interval=20 where interval changes the frequency of logging https://your_app_name/script?statistic=cpu logs CPU usage instead of memory

Better Output

Output as a downloadable log file. Output as a page in the web app. Graphed output

mdsauce/navi