KubeCon + CloudNativeCon India 2025

Expedia Group’s RCP platform is transitioning to ArgoCD as its control plane. As part of this journey, we are adapting ArgoCD’s industry-leading technology to meet our current and future needs. A critical component of this effort involves fine-tuning the repo-server, originally optimized for GitOps workflows, to manage over 4,000 unique Helm charts at a scale of 12,000 applications, with a forecasted future scale of 30,000 applications.

This talk delves into our deep-dive scalability testing of the repo-server component, exploring memory usage patterns, troubleshooting performance bottlenecks, and addressing memory burst issues. We share our solutions to stabilize the repo-server under heavy load, and provide actionable recommendations based on configurations and parameters that worked for us. These insights aim to benefit the broader community in their own scalability endeavors.

Kueue and Volcano are leading the charge in orchestrating large-scale distributed ML jobs. But are they truly maximizing your GPU resources? Traditional batch scheduling methods often suffer from inefficient queue management, and rigid allocations that fail to adapt to real-time demand resulting in problems that scale with workloads.

This talk dives into how priority-aware queueing and elastic resource allocation can supercharge Kueue and Volcano, making batch scheduling more adaptive and efficient. We’ll break down the scheduler’s architecture, exploring how jobs dynamically move between priority queues, how elastic scheduling adjusts resource allocations in real time, and how these improvements lead to faster job execution and better GPU utilization.

Whether you're managing distributed training, hyperparameter tuning, or large-scale inference pipelines, this talk will provide the tools and strategies needed to unlock smarter scheduling and maximize ROI on Kubernetes GPU workloads.

In Kubernetes, DNS resolution is vital for service discovery. CoreDNS, the default DNS provider, can face performance issues such as timeouts, latency, and scalability challenges, which increase reliance on upstream DNS and lead to unreliable service discovery. To enhance DNS reliability, we propose integrating Node Local DNS Cache.

We'll discuss implementing node-local DNS caching in 270 clusters with over 18,000 nodes, effectively resolving DNS issues, boosting resolution reliability, reducing CoreDNS load, and minimizing dependency on upstream DNS sources. Attendees will gain insights into the technical approach, including performance assessments and testing scenarios emphasizing latency improvements. We've achieved a 30x performance enhancement and greater dns query-handling capacities through node-local DNS cache deployment. This session will provide valuable understanding of techniques to ensure robust DNS resolution in large-scale Kubernetes environments.

As organizations shift workloads away from centralized clouds, they need solutions that are both lightweight and scalable. In this lightning talk, we’ll see how k0s - a single binary Kubernetes distribution, bridges this gap by making edge deployments simple and efficient.

We’ll walk through a short demo on Raspberry Pi 5, showing how to set up a worker node using a Pi device and deploy containerized applications closer to where data is generated. By the end, we’ll understand how this minimal approach can power IoT, AI/ML.

Jumping into a massive infrastructure codebase as an intern can be intimidating. But here’s what made onboarding smoother: unit testing with Kubernetes APIs and fake clients.
This talk shares my experience of working on unit tests for an infra plumbing component of a data lakehouse.

The journey from being overwhelmed by Go and Kubernetes to realizing that unit testing is the best entry point to understanding large-scale cloud-native systems will be explored.

The session highlights why Kubernetes APIs make testing easier, introduces informers, controllers, and API behaviors and how they help in testing, how fake clients eliminate the need for a real cluster, and why this is a great starting point for interns and students working on their first cloud-based project.

Attendees will gain insights into the advantages of Kubernetes API interactions and testing, including speed, reliability, comprehensive coverage, and improved developer experience.

Kubernetes administrators often struggle with maintaining consistency across distributed nodes, managing OS updates, and preventing configuration drift in large-scale containerized environments. Wouldn't it be great if there is a way to generate the distribution using a container file like we do for application images? With bootable containers (bootc) you can do it.

In this demo-centric presentation, we will explore how bootc can simplify operating system management by treating itself as an immutable container image. Using a local Kind cluster on Podman Desktop, we will demonstrate how to create and deploy images, showcasing its advantages in maintaining consistency and security across clusters and edge nodes. By treating the OS as a container image, we can take advantage of existing container workflows, making updates and rollbacks atomic and version-controlled.

Attendees will learn how they can use their container-building skills to manage their distro images as well.

Do you know about Large Language Models or LLMs? It’s been a while since we’ve seen a technology like this grab everyone’s attention, techies and non-techies alike. Ever since OpenAI’s ChatGPT burst onto the scene, powered by an LLM, AI has become a household name. But for newcomers, especially in the cloud native ecosystem, these concepts can feel overwhelming.
In this talk, we’ll start from the very basics, explaining what LLMs are and busting some common myths. We’ll break down the key terms with simple diagrams and analogies that are easy to digest. You’ll get a clear, beginner-friendly look at how LLMs work, why Kubernetes is perfect for running them, and how they fit together in the cloud-native world.
We’ll even walk through running an LLM on Kubernetes using vLLM for some practical insights and touch on how this opens the door to AIOps. If you’re curious about stepping into the AI ecosystem and understanding its cloud foundations, this talk is for you.

Running a homelab means running a diverse set of applications like NextCloud, readeck.org, usememos.com.

They are written in many languages and frameworks and in general lack a good way to monitor them. The old way to monitor them involved having a basic blackbox_exporter enabled synthetic monitoring. Even instrumented, each application has a different set of metrics, leading to complex alerting and dashboards.

Grafana Beyla, powered by eBPF and OpenTelemetry, combined with Prometheus promises to generate first class observability signals for services regardless of the language and frameworks used. We will walk through deploying Beyla for a homelab, and the challenges in having good observability even with Beyla.

We will walk through:

1. Deploying Beyla in your environment with Prometheus
2. Setting up dashboards for the data
3. Robust alerting over the signals, especially in low, sporadic traffic (homelab) scenarios.
4. The missing signals from eBPF and how you can fill those gaps

Kubernetes Operators are revolutionasing cloud-native application management—but when designing an operator, should we go one CRD per operator or manage multiple CRDs within a single operator? How does cascading CRDs impact reconciliation, dependency handling, and performance?

In this talk, we'll dive deep into real-world use cases to explore the trade-offs between these approaches. Whether we are building custom Kubernetes controllers or optimising cloud-native infrastructure, this session will help one make the right design choice for scalability, maintainability, and performance.