Yair Etziony
Yair Etziony is an Engineering Manager at AMBOSS with over 30 years of experience in systems administration, DevOps, and cloud technologies. He holds a BA in modern history and philosophy, organizes the Berlin DevOps meetup, and has spoken at DevOpsCon and DevOps Days Berlin. His career spans from VAX/VMS systems in the 1990s to modern cloud infrastructure, providing a unique historical perspective on technology evolution.
Sessions
Day One operations focus on initial deployment, provisioning resources, and achieving operational state. Day Two operations encompass ongoing lifecycle management: updates, drift correction, decommissioning, and sustained operational health.
The evolution from Puppet and Chef to Ansible, Terraform, Pulumi, and platforms like Crossplane represents decades of innovation in infrastructure automation. Each generation has advanced deployment capabilities, yet consistent patterns emerge in post-deployment operational challenges.
This talk explores the intricate transition from deployment to sustained operations, a journey marked by recurring themes in the evolution of configuration management. We'll dissect how different approaches handle this transition, exploring operational challenges across infrastructure layers from bare metal through hypervisors to cloud platforms.
Through comprehensive historical analysis, we'll delve into questions such as: Why do similar operational challenges persist across different generations of tooling? How do state management approaches compare to distributed coordination systems? What can we learn from examining infrastructure as dynamic systems rather than static code?
Additionally, we'll explore the relationship between infrastructure automation tools and observability platforms, investigating how the separation between these domains affects operational visibility and decision-making during Day Two operations.
You'll gain insights into approaches to infrastructure management that address the complete operational lifecycle, leaving you prepared to tackle the challenges ahead.
After 30 years in this field and too many years away from actual code, I came back. What I found frustrated me enough to start building. AI gave me the velocity, but the direction came from decades of seeing the same patterns return, dressed in new syntax.
I built most of these tools around Kubernetes. Not because I love it, but because I was already frustrated from working with it. After five years of managing a DevOps consultancy, I understood something: you can't convince everyone who doesn't need it not to use it. It's a standard now. It's everywhere. So I stopped fighting.
But going deep into its ecosystem, really learning it, something clicked. I remembered there was something else out there. Something that could do some of these things more elegantly.
Eleven tools in seven months. Still a work in progress. Our modern stack has real engineering in it. But somewhere along the way, we fragmented what used to be unified. We solved problems that were already solved, just in places we weren't looking.