I present now the case against CI/CD:
- It’s expensive to set up CI/CD pipelines, and the cost of doing so is not paid back
- It’s expensive to maintain CI/CD pipelines so they keep working – this is not adding value to the software
- How do I justify the provision of capacity to release several times a day, compared with our usual quarterly release cycle?
- The time spent on build scripting could be more effectively be spent building and debugging the software
- We can’t release frequently because we need to test the software first and that means the investment in pipeline never pays off
Let the judges and jury retire…
It’s Reversing the Burden of Proof
Consider many real world supply and demand scenarios:
- The waiting time at the supermarket is usually only a couple of customers
- The postman visits daily, rather than coming monthly with a month’s worth of post
- Your coffee is delivered while it’s still hot, rather than in an hourly release cycle
Why?
The default process for supply and demand is that when something’s ready it’s shipped.
But my features take months to get ready – that’s because your batch size is very high and you’re not working incrementally. If there’s indivisible value in things you do and they HAVE to take that amount of time, then sure… it’s an exception… but it’s rare and probably not actually the case.
But our testing needs to be thorough – yes, so why haven’t you automated it to run quickly and give you instant feedback the moment something goes wrong?
But there’s risk in constantly changing stuff – there is, but the risk is lower in a low batch-size change, than in a higher.
What’s the Point of CI/CD?
- Encourages some great low-risk practices like test driven development and small, incremental batches
- Makes builds repeatable as they’re done by machine
- Encourages “no voodoo” in the release process, making it less likely to go wrong
- Makes a home for all the build, test and release automation to be orchestrated
What’s the Point of Frequent Release?
Every valuable thing we make only has real world value once it’s live. Until then it’s festering with sunk cost but no return. It runs the risk of being broken, or forgotten, or missing its market opportunity.
Conversely, going live as soon as practical gives return on investment.
The fact that there’s not a quarterly release cycle has the virtuous feedback loop of encouraging smaller requests as people don’t have to save up their “orders” for the big drop.
Example, my shopping cart is small when I’m dropping by the local shop, but if I lived in an isolated island and could only order shopping for a monthly air drop, then I’d have to over supply “just in case”.
Where’s the Value in CI/CD?
Non CI/CD is a business cost. “Owing to development practices we can’t safely release our software too often, so you’ll just have to wait…”.
No business wants to be held to ransom for their features.
Conversely, CI/CD enables continuous flow to live.
But I Have Genuinely Long Feedback Cycles
Try not to…
But if you do, the CI/CD techniques can be run in the slow lane for higher environments, while the latest code going into the dev environment, using the same tools as will ultimately put them live, are giving continuous feedback on the code that WILL go into the release train in future.
But…
No.
Published on Java Code Geeks with permission by Ashley Frieze, partner at our JCG program. See the original article here: The Case Against CI/CD Opinions expressed by Java Code Geeks contributors are their own. |
I think you are better at standup comedy than software development. For your own good, keep to the former!
This article seems to be written from the perspective of not having experienced the benefit of modern practices. Please seek help.
I’d genuinely like to spend some time demonstrating what you seem to be missing out on, here, but you should really be able to keep up with industry standards and revelations. You can’t, evidently, and you’re arrogant about it, too.
Calm down gang. I’m completely in favour of CI CD. Check out my other posts on the original site for context. I am making a counter argument I’ve read about and disagree with.
Note. The majority of this article is in favour of CI CD. Read past section one! I too was shocked to read people making the former arguments. In truth, reality has nuance, but it’s a good idea to start with CI CD when a project is still small and easy to control.