I'm currently working at a startup that's been around for two years, has a solid customer base, and a development team of about 8 people. We have a typical web app built with Spring Boot and React, along with a number of LLM interactions and data management components. The admin app is still using a mostly manual git workflow, and there are no unit or integration tests in place; we're just starting to implement some selenium testing. Our regression testing takes a whopping 5 days, and deployment is quite primitive, relying on shell scripts without any CI. Our infrastructure uses Terraform, but it's out of date with significant config drift. We operate across several AWS environments including dev, QA, pre-prod, and production. At this stage of development maturity, I'd love to hear how others approach automation in similar situations. What should I focus on first?
8 Answers
In my experience, it's all about focusing on where the pain points are. Listen to your developers and find out what tasks are wasting their time or preventing them from pivoting quickly. Automate the things that light up as inefficiencies, even if it seems like a short-term fix. Startups often need rapid change, so don’t hesitate to prioritize speed over perfect solutions!
Updating Terraform should be high on your to-do list, and I'd recommend developing an end-to-end suite that you can hand off to your developers. This should help ensure more reliable releases.
A good approach is to start automating from the ground up, especially if it's a greenfield project. You might find it more efficient to put in automation before you even run into issues for the first time.
Make sure security is your top priority here. In a startup, vulnerabilities can easily lead to major issues. Get your infrastructure patched up and ensure processes are in place to keep everything current, then focus on reliability and reproducibility.
Agreed! But also take a moment to consider the engineering culture; the lack of tests after two years indicates bigger issues that need addressing, like team discipline and leadership.
Just get started! Focus on automating what you can early on, especially where you're feeling overwhelmed. Look for those tasks that are easy to automate that will save you time in the long run.
Definitely start by addressing the main pain points. If the regression tests are taking 5 days, that's a huge bottleneck! Help them ramp up automation, especially with your selenium tests. Getting some one-button deploys in place could be a game changer. Also, make sure you implement observability once you start deploying more frequently, as it will help with tracking issues down the line.
You typically want to automate tasks that you expect to repeat often, or tasks that are risky when done manually. With no automated tests in place and that long regression period, I'd say you definitely have a lot to work on! Shell scripts might work for now for deployments, but think about introducing more automation in that part too for long-term efficiency.
First things first, get your Infrastructure as Code (IaC) and Continuous Integration/Continuous Deployment (CI/CD) pipelines sorted out. This will save you so much hassle down the road!
Absolutely! Getting that all set up in a CI/CD pipeline will streamline everything.