joakimk

Background

We started out with two apps.

• foo-staging
• foo-production

Our CI would run tests, then deploy to staging, run smoke tests and then run to production and run smoke tests.

This took about 22-25 min.

joakimk

If you're seeing this error, try heroku ps:scale web=0 then import the database again and run heroku ps:scale web=1 afterwards. Found nothing on google for this error, so posting this gist with the solution that worked for me. Could have been a temporary error too.

joakimk

Some notes on my Elm development environment.

The most important thing is elm-format. It auto-formats the code when you save. You will likely miss this in other languages.

0. brew install elm elm-format
1. Install Atom
2. Elm plugins: elm-format language-elm
3. Vim plugins: ex-mode vim-mode-plus

Currently evaluating the elmjutsu plugin.

joakimk

require 'rubygems'
require 'httparty'
require 'nokogiri'

# This tool removes old builds from queues so that only the latest commit gets run.
# This way slow jobs won't fall behind, they will just skip commits upto the latest
# available commit that was successful in upstream projects.
class JenkinsQueueOptimizer
  class Api
    include HTTParty

joakimk

/* jshint asi:true */
// PixiJS v4 example (WebGL rendering)

model = loadStateOrDefaultTo({
    move: 1
})

tick = (delta) => {
    if(codeHasChanged()) { return }
    

joakimk

Instant-feeling unit tests improves productivity and developer happiness :)

Elixir tooling isn't quite there the way Ruby can be when optimized.

Ruby benchmarks:

• instant 46ms base time (time ruby -e "")
• instant 190ms running a ruby unit test doing nothing
• in 50k line rails app:
• okay 640ms bundler (time bundle exec ruby -e "")

joakimk

http://privpakportal.schenker.nu/TrackAndTrace/packageinfo.aspx?packagesurrid=69682901&inpPackageId=<SSCC NO>

joakimk

require 'socket'

class Graphite
  def initialize(host)
    @host = host
  end

  def socket
    return @socket if @socket && !@socket.closed?
    @socket = TCPSocket.new(@host, 2003)

joakimk

I'm working on something that will be way better than using jenkins to do continuous deployment, but here is how we do it today.

This is in reply to https://twitter.com/#!/haraldmartin/status/185683988063588352

What we want: For each revision, build it in steps, don't advance if the previous step failed. So that it can run a chain of steps like: unit_tests -> integration -> deploy staging -> deploy production.

If you just join projects in jenkins, and say rev-a passes in project-a, project-b might get rev-b. That's not good if project-b happens to be deploy-to-production and rev-b is broken.

We use: https://wiki.jenkins-ci.org/display/JENKINS/Join+Plugin And: https://wiki.jenkins-ci.org/display/JENKINS/Parameterized+Trigger+Plugin

joakimk

require 'faye'

Faye::WebSocket.load_adapter('thin')
use Faye::RackAdapter, :mount => '/faye', :timeout => 25

require ::File.expand_path('../config/environment',  __FILE__)
run App::Application