I'm part of a site reliability engineering (SRE) team of 12 at a car rental company, and we've been tasked with coming up with ideas to incorporate AI tools into our project. We manage a wide range of environments, primarily hosted in AWS, where we oversee around 1200 servers across different setups, including some using EKS and ECS.
Currently, we're handling various tasks like Bitbucket administration, Terraform for managing infrastructure, troubleshooting Kubernetes, and managing Jenkins pipelines. We're also using tools like ServiceNow, Jira, and Confluence for documentation and ticketing.
I'm particularly interested in finding ways to implement AI to help with the challenges we face in Kubernetes management, as many team members struggle with troubleshooting. If anyone has experience with successfully integrating AI into similar tools or projects, I'd love to hear your thoughts!
4 Answers
Consider using AI to parse build and deployment logs! If app developers occasionally merge problematic code and break the pipeline, you could have AI analyze the logs, identify errors, and offer specific guidance. This way, when the pipeline fails, the alerts could include actionable fixes rather than just links to the failure details. That could really expedite troubleshooting!
One cool idea could be to implement a chatbot that helps with documentation. Since you have such a variety of environments, making documentation more accessible could streamline developer workflows and assist with troubleshooting. Just frame it as a way to make things easier for the team! I'm not a fan of complicating other systems with AI, though.
For Kubernetes, we experimented with using GPT models at Microsoft to analyze `kubectl` outputs and suggest common fixes. It can be effective if you start small—maybe a Slack bot that takes pod crash logs and suggests potential causes. Keeping the project scope narrow helps, especially since K8s errors tend to follow certain patterns. Just remember to supervise its suggestions, like a junior engineer!
I recently had AI generate a Terraform script for over 800 resources—it was super handy! It worked well using Copilot in VSCode. I also had AI write a shell script to tackle a common Kubernetes issue, and I set it up to run in a GitHub action. Plus, I'm planning to have AI help me document my code after it's complete. It's a great way to utilize AI for repetitive tasks!
Exactly! It could save lots of time for developers who are often left guessing what's wrong after a failure.