Bart Farrell: First things first, who are you, what's your role, and where do you work?
Mesut Oezdil: I am Mesut. I am a DevOps engineer at Adfinis.
Bart Farrell: A Kubernetes setting can look wrong but still feel risky to change once it's already in production. Requests, limits, auto-scaling, or probes. What would you tell a team that sees the problem but is nervous the fix could cause an outage?
Mesut Oezdil: There are four steps for it. Number one, you should check the metrics and logs. Observability plays a great role in this stage. Number two, you should change just one thing at a time. Isolation is a great thing here. Therefore, I will say change just one thing. Number three, test all the things in your test staging, not in production or something like that. And last thing, be ready for rollback. That's the solution with four steps.
Bart Farrell: We're here at KCD Helsinki. AI is a topic of conversation, but is it something that you're actually using on Kubernetes, or do you think we still got a ways to go?
Mesut Oezdil: It is a huge topic, maybe for four months we have been discussing about AI for Kubernetes. There are some options like from Solo.io, kagent, kgateway, etc. And it is an amazing option for using Kubernetes in your main production. I'm a contributor of kagent and I see there is a great chance and we can use such things in our production in our Kubernetes environment but there are always security problems and we have lots of issues and it is not easy to decide such things in your production.
