tl;dr
I attended the DevOpsDays in Graz - the first DevOpsDays in Austria with around 100+ attendees. Day 1 had workshops at FH Campus 02, Day 2 talks at WKO Europasaal. Highlights: Rainer Stropek's Azure OpenAI workshop with Bicep, Daniel Radl on vendor-agnostic cloud solutions, and Marko Bevc on Terraform secrets.
DevOpsDays finally came to Austria with the first edition in Graz. Since it's only about 2 hours from Klagenfurt and the topics looked interesting, I decided to visit it and see how the community is and get inspired.
The conference was on September 5-6, 2025, with something above 100 attendees (no official numbers published, but that's how it felt). First day workshops, second day talks.
Getting there
I drove from Klagenfurt this time, about 2 hours by car. There are bus connections available, see my Linuxtage post for more details.
Venues
The conference was split between two venues:
- Day 1: Workshops at FH Campus 02
- Day 2: Sessions at WKO Europasaal
Both locations were easy to find and close to each other, though there are some construction sites around that can make it tricky when visiting for the first time. Pretty good setup for the event overall.
At the end of the first day a heavy storm hit Graz. I heard that this caused flight delays for people coming from other countries. For me it was a pretty wet walk to the apartment.
Day 1
The first day was reserved for the workshops, and there were multiple workshops running in parallel. A registration form for the Workshops was sent before the event, so you had to decide for one of the Workshops in the morning, and one in the afternoon. Here are the ones that I selected:
Pulumi - an IaC Tool Using General-Purpose Languages
This workshop was a bit short - only used about half of the time slot, but we got a good hands-on introduction to Pulumi's approach. Unlike traditional IaC tools that use domain-specific languages, Pulumi lets you write infrastructure code in familiar programming languages like Python.
Lisa Gößler and Valentin Klammer-Krizaj started with the basics: setting up a local Pulumi backend (pulumi login --local) and creating a new Kubernetes project with Python. The workshop used minikube as our local Kubernetes cluster for testing deployments.
The workflow felt familiar yet different from Terraform. We used pulumi preview to see planned changes and pulumi up to apply them. One interesting aspect was learning about Pulumi's approach to resource lifecycle - it distinguishes between delete/create cycles versus in-place replacements, which can be important for production deployments. A key advantage highlighted was Pulumi's asynchronous execution model. You can add callbacks to work with provider results, giving you more programmatic control over complex deployment scenarios.
While we just scratched the surface, it was enough to see Pulumi's potential for teams already comfortable with general-purpose programming languages. The ability to use existing Python libraries and testing frameworks alongside infrastructure code is compelling, though it does come with a steeper learning curve compared to other tools.
Learn by Example - Deploying and Securing Azure OpenAI
Rainer Stropek ran this workshop and it was very good. He showed some very cool stuff related to using OpenAI models with Azure, focusing on practical deployment and security considerations.
The workshop centered around building a RAG (Retrieval-Augmented Generation) system using Azure's AI services. Rainer highlighted key challenges when working with AI in enterprise environments: model cutoff dates mean your AI doesn't know about recent company developments, and you need strategies to inject current organizational knowledge into responses.
He demonstrated Azure's advantages for European organizations, particularly the ability to choose specific datacenters for inference - important for data and compliance requirements. The architecture combined Azure OpenAI (AOAI) with Azure Search as a vector database, showing how to build a complete AI-powered search and response system.
An interesting security topic was the concept of "AI Wallet Attacks" - where malicious users exploit AI services through applications to run up costs or extract unintended information. This reinforced the importance of proper authentication, rate limiting, and monitoring when exposing AI capabilities.
He chose Bicep for all the infrastructure provisioning, which made perfect sense for this use case. Since we were already locked into Azure-specific services (Azure OpenAI, Azure Search), using Azure's native IaC tool was a nice way to see something new in action. Bicep scripts are idempotent and don't require state management like Terraform, making them simpler for Azure-only deployments.
Day 2: Sessions
Political Deployments and EU Regulations
Kerim Satirli from HashiCorp opened with a keynote about the regulatory mess we're dealing with in Europe. GDPR, NIS2, the Cyber Resilience Act coming in 2027 - it's a lot. His point about needing unified EU regulations instead of 27 different interpretations made sense. No system is perfectly secure anyway, and vendors oversell their solutions.
Paulina Dubas later covered more of the EU regulation landscape - AI Act, Data Act, DORA. The takeaway was that logging, monitoring, and governance are becoming compliance requirements, not just nice-to-haves. Her line "Our code is more legal than ever" summed it up well.
GitHub Automation at Scale
Eugen Bozic shared real experience automating GitHub Enterprise for 1,600 users and 1,400 repos using Terraform. The honesty was refreshing - he talked about the pain points, not just the wins. API limitations, state management issues, the need for dedicated maintenance. His advice to split Terraform configs into smaller segments and start small seems very practical.
Cloud-Native Yes – Vendor Lock-In No!
Daniel Radl gave this talk and I really enjoyed it. He did a live demo deploying a chatbot on STACKIT, showing the full flow with Terraform provisioning and deploying to Kubernetes with Harbor.
What I liked was his honest evaluation: Is it cloud-native? Yes. Vendor-agnostic? No - because we're using STACKIT-specific services. Developer-friendly? Yes, deployment was fast. This highlighted the trade-off between data sovereignty (important for EU) and true portability.
Safely Storing Secrets in Terraform
Marko Bevc presented this one and it was insightful. I discovered some new ways to deal with secrets in Terraform, especially when it comes to state files. His memorable quote captured the security mindset perfectly: "Dance like nobody is watching, but definitely encrypt like everybody is."
The key new thing for me was ephemeral resources - they let you handle secrets without storing them in state files. Combined with write-only attributes, it's a much better approach than what I've known before. He covered the full multi-layer approach: encrypted state, secret managers (Vault, AWS Secrets Manager), proper access controls, and team education.
Networking
I had great chats about AI with Stefan Ellmauthaler on the first day. During lunch on day 2 (which was a bit off schedule due to technical issues), I talked with Marko Bevc and Marko Zabreznik. They convinced me that I need to attend DevOpsDays Ljubljana as well.
There was enough time between workshops and sessions for discussions. On day 2 there were also sponsors, and I had some very interesting discussions with Dino Begićević from NTS and Stefan Vucak from BearingPoint about the current state of IT, the evolution of DevOps, and AI.
Before I left, I also chatted with Daniel Radl from ebcont about AI topics and data sovereignty and the benefits of STACKIT.
Organization
Everything went well overall. The only issues were some technical delays on day 2 that caused schedule problems. Since it was a single track, some participants who had to leave earlier might have missed sessions. Day 1 on the other hand went on schedule.
As typical with Austrian conferences, people are not as open to talk as at other conferences. I hope this will change with future editions. The organizers should emphasize the Pac-Man Rule at the beginning.
Conclusion
It's a small event and was the first edition, but it shows promise for Austria's DevOps community.
I may play around with Pulumi and will look into STACKIT. The conference reinforced that Azure, Terraform, and Kubernetes are still the main topics in the DevOps space.
It was a nice experience - took some new topics with me and look forward to the next edition.