Kubernetes Is My New Boss

---

September 26th, 2016. It’s been a busy year already, and Kubernetes is still riding high after the container wars—like a triumphant general returning from battle. I’ve found myself spending more time with it than I care to admit, especially on days when things just aren’t working as they should.

Let me start by saying that I’m not one of those early Kubernetes enthusiasts who saw its potential immediately. When we first started using it at my company, it was a bit of a struggle to get everything set up and running smoothly. But over time, the benefits have become undeniable. Today, Kubernetes is practically like an assistant in charge of my day-to-day engineering tasks—helpful but sometimes difficult to deal with.

The other big player this month has been Helm. We’ve been experimenting more deeply with it, using it for managing our Kubernetes clusters and deploying applications. The promise of “reusability” from charts is exciting, but the learning curve can be steep. Our team had a few rough mornings figuring out how to package and deploy complex services effectively.

Then there’s Istio, which has been getting a lot of attention lately. I’ve heard a lot about service mesh and how it could transform our microservices architecture. It seems like everyone is either already using or planning to use Istio, but we’re still holding off for now. The complexity around sidecars and the overall setup has made us wary.

Serverless hype hit its peak this month with AWS Lambda announcements, and I’ve been keeping an eye on it. I’m not entirely convinced that serverless is right for every problem yet—there’s still a lot of friction around cold starts and cost optimization. But I can see where it might make sense for certain parts of our stack.

Back at the ops desk, we’re seeing more use cases for Terraform 0.x. The move to version control for infrastructure as code has been gradual, but now that we’ve got a few pipelines in place, it’s hard to go back. However, there are still some rough edges—issues with state management and remote execution bugs that keep popping up.

The term “GitOps” started making its rounds this month, which is kind of hilarious because I’ve been practicing something similar for years without even knowing it had a fancy name. We use Jenkins to sync our Kubernetes manifests from source control, but there are still moments when things break and we end up doing manual rollbacks.

On the monitoring front, Prometheus + Grafana have become indispensable tools. They’re incredibly powerful, but also somewhat overwhelming at times. It’s not uncommon for us to spend hours debugging strange spikes or trying to understand why certain metrics aren’t behaving as expected.

I’ve been spending some time playing with Envoy recently too—partly because I’m curious about the performance benefits, and partly because it seems like a great fit for our service mesh needs. The learning curve is steeper than I’d hoped, but the potential gains are significant if we can get everything working right.

Overall, this month has been filled with a mix of excitement and frustration as we continue to integrate these new technologies into our infrastructure. Kubernetes still feels like it’s got more features than I know how to use, Helm is a black box that sometimes just doesn’t work, and Terraform is an ongoing battle against its own complexity.

But here’s the thing—despite all the problems, I wouldn’t have it any other way. The tech world moves fast, and if we don’t embrace these new tools and techniques, we risk getting left behind. As always, I’m just trying to figure out how to make sense of it all while keeping my sanity.

Stay tuned for more adventures in Kubernetes and beyond!