Good question! I think a lot of devs get therapeutic enjoyment from setting up CI. It starts out as fun for the first few hours, while everything is new and exciting. However, it turns to disdain very quickly when:
- you need to remember how to set up postgres/rvm/lucene again
- you need to add a 2nd server and realize you now need to manage a cluster (and you didn't automate the first one!)
- you're away on vacation with your family, and someone calls you because the company CI server is down
- you realize the tests run too slowly and now need to figure out how to make them faster
- you get a bad performance review because of all the valuable engineer time you're wasting managing the CI server, which you could have outsourced for a fraction of the price
- you wonder how you turned into "the build guy" when you actually signed on to make great products.
These are the sort of problems that hosted CI like https://circleci.com solve for you :)
This is a general problem with any kind of maintenance or setup that you might do. Did you write a diary of how you went about setting it up ? Did you communicate to the rest of your team how it's supposed to work / how you suspect it might break ?
If not, don't be _too_ surprised to be The One Person who gets called about that thing.
Being "the build guy" is an important role in delivering great products. In fact, SaaS CI offerings in my experience cater to a small subset of real world apps, none of which are usually heavily used non-trivial apps. I helped build one SaaS CI product and really hope this changes in the coming years as we abstract more 'things'. But for the time being, 'build guys' or DevOps guys or whatever the latest word is play a vital role.
That's why sysadmins should be encouraged to never configure systems directly, but use write-apply-observe-refine cycle of Puppet/Chef (or alike) declarative configurations.
Sounds like a hassle (and sometimes it is, especially in cases where to-be-configured software has Turing-complete configuration files), but in my own experience actually proved very helpful.
There's quite a bit more than the money involved when setting up SaaS infrastructure; even if the first 3 months were _sponsored_, the precious time invested would be wasted if they aren't signing up for the long haul.
And if they are signing up for the long haul, then 3 months of X discount makes very little difference in the calculation.
If your tiny startup isn't able to afford a couple of hundred dollars a month for various hosted services after 3 months, then you're prioritizing wrong. You need to focus on building a valuable product and getting it to customers, not on building your own internal setup to save a little bit of money. Companies doing the former are much more likely to succeed than those doing the latter.
Thanks for putting up the more relevant comparison price point, though, and I do agree with you that outsourcing non-core functions is essential to getting real work done on the actual product.
However, am I the only one who enjoys setting up Graphite and CI and similar software? (No email, though. Did it once, and don't want to give up sleeping time for a week to do it again.)