How I learned to stop worrying and love the unicorn...

blog.teachstreet.com

37 points·swindsor··5 comments

Great article, and fun to see what is going on inside. We used a similar strategy with great success at my last company (though we were using Passenger).

Once you've nailed the zero-downtime deployment, it's a short hop to adding a small box (or a script on your test server) to do continuous integration and then moving from there to continuous deployment. For our small team this setup was really liberating.

If you head down that path, I'd love to see updates on what your team's experience is with that transition.

We've stop/started continus integration a few times with a few different CI servers with mixed success. What did you guys end up using?

Right now, we have tests, but I'd still say we rely pretty heavily on hoptoad to let us know if we really broke something badly. ;)

Yeah, we relied on hoptoad as our backstop too. We used CruiseControl.rb running on a cheap low end server with a plugin for doing deployments.

https://github.com/thoughtworks/cruisecontrol.rb https://github.com/aentos/ccrb_cap_deployer

The code is simple to understand, self contained, based around rake, and worked almost immediately for us (after some configuration). We were a two person team. I don't know if the setup we had would have worked with a larger team.

It felt like it made an impact in our ability to turn code faster. We had a solid test suite, which helped.

Nice writeup -- at Jambool/Social Gold we used a similar approach which enabled us to do deployments continuously throughout the day without interrupting transactions. This was key to allowing us to iterate faster and incrementally deploy small pieces of larger projects without having take downtime.

Going from 1-2 deployments a day to 10 without downtime - Awesome stuff!

Reply on news.ycombinator.com