2.0 -//Pentabarf//Schedule//EN PUBLISH DSKA8T@@talks.devopsdays.org -DSKA8T It's too late to scale up now: how to choose the right metric to scale en en 20261113T150000 20261113T170000 020000 It's too late to scale up now: how to choose the right metric to scale CPU and memory are the defaults for scaling up. But in many projects, these metrics react minutes late to issues caused by other areas of the application, thus delaying your autoscaling. In this workshop, we will learn how to hunt for the right metric and how to scale once you find it using Kind, Keda, and Prometheus. PUBLIC CONFIRMED Workshops - 2hrs https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/DSKA8T/ Bogatell Workshops (English) Christian Polanco PUBLISH DER3ET@@talks.devopsdays.org -DER3ET Intro to Value Stream Mapping for Leaders en en 20261113T173000 20261113T193000 020000 Intro to Value Stream Mapping for Leaders In this hands-on session, you'll map and measure a real delivery workflow, identify bottlenecks, and apply lean principles to improve your delivery's speed and quality. The workshop distills 10+ hours of mapping exercises and two years of mapping experience into a practical, do-it-yourself format. Come with a delivery process in mind; leave with a concrete map and an action plan. Agenda: - Arrival and introductions - Overview of value stream mapping - Hands-on mapping exercise - Group discussion on constraints & solutions - Action planning & next steps PUBLIC CONFIRMED Workshops - 2hrs https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/DER3ET/ Bogatell Workshops (English) Miguel Dias PUBLISH VYEFPG@@talks.devopsdays.org -VYEFPG Getting started with OpenTelemetry instrumentation: A practical playbook en en 20261113T150000 20261113T170000 020000 Getting started with OpenTelemetry instrumentation: A practical playbook OpenTelemetry has become an essential open source framework for standardizing telemetry data. In this beginner-friendly hands-on lab, you’ll walk through the journey of instrumenting an application with OpenTelemetry and sending metrics, logs, and traces to the Grafana LGTM stack (Loki for logs, Grafana for visualization, Tempo for traces, and Mimir for metrics). You’ll learn common instrumentation approaches, such as adding application-level telemetry using the OpenTelemetry SDK, collecting basic infrastructure metrics with the OpenTelemetry Collector, and implementing cross-service communication with zero-code eBPF instrumentation. You’ll also apply practical customization techniques, such as enriching and sampling spans, tuning metrics, and adding your own signals when auto-instrumentation isn’t enough. By instrumenting a sample application step by step, you’ll leave this lab with a clear roadmap for choosing the right OpenTelemetry instrumentation approach within your own environment. By the end of the three hours, you will: - Have a core understanding of the OpenTelemetry project and its key components - Select the right level of instrumentation based on various use cases - Understand how zero-code eBPF instrumentation helps reveal cross-service communication - Instrument applications using OpenTelemetry SDKs - Apply end-to-end instrumentation patterns to get started quickly and improve over time - Send telemetry to the Grafana LGTM stack and understand how it’s used - Customize telemetry to manage cost and noise Requirements: - Laptop with WiFi access - Linux/MacOS/WSL2 - Docker - Docker compose - Administrator or sudo access to build and run Docker containers - We will provide a docker-compose file containing what is needed for this lab. This session is tailored for developers and SREs. PUBLIC CONFIRMED Workshops - 2hrs https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/VYEFPG/ Pallars Workshops (English) Mario Macias Miguel Angel Ortega PUBLISH KAXXTX@@talks.devopsdays.org -KAXXTX Self-Healing AI on Kubernetes: Build an Operator for Your LLM App en en 20261113T173000 20261113T193000 020000 Self-Healing AI on Kubernetes: Build an Operator for Your LLM App Running an AI app on Kubernetes usually means a pile of Deployments, Services, ConfigMaps and Secrets that someone maintains by hand. An operator replaces all of that with one declarative resource: you describe the agent you want, and the cluster keeps it running, configured, and healed. In this workshop you'll build that operator yourself. Starting from a provided skeleton, you'll define an AIAgent custom resource (its model, replicas, and config) and implement the controller that reconciles it: an informer that watches your AIAgent objects, a work queue, and a reconcile function that creates and maintains the underlying Deployment and Service running a lightweight local model. Then we make it earn the name "self-healing" — delete the agent's Deployment and watch the operator recreate it, bump the replica count and watch it scale, change the model field and watch it roll out. The sample agent runs on CPU against a local cluster (no GPU, no external API keys required), so everyone gets a working AI app on their own laptop. We'll close by mapping what you built onto how production AI operators handle the harder parts — GPUs, model storage, autoscaling. PUBLIC CONFIRMED Workshops - 2hrs https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/KAXXTX/ Pallars Workshops (English) Kritik Sachdeva PUBLISH N9BX9T@@talks.devopsdays.org -N9BX9T Tu sistema no es lento, lo estás midiendo mal es es 20261113T150000 20261113T170000 020000 Tu sistema no es lento, lo estás midiendo mal Most performance issues are solved the same way: add more CPU, more memory, or scale out. It works... but it’s often an expensive shortcut that hides the real problem. In this session, we’ll explore a practical approach to performance engineering using open-source tools and reproducible methods. Instead of relying on intuition or single metrics like CPU usage, we’ll break down how to characterize workload behavior (CPU, memory, and I/O patterns), identify system limits through breaking-point testing, and measure efficiency using cost per unit of work. Through different examples, we’ll analyze how multiple workloads behave under load, why average latency is misleading, and how to detect the “knee of the curve” where systems begin to fail. We’ll also dive into hands-on optimization techniques (including memory-efficient patterns, caching strategies, and SIMD vectorization) showing how code changes can outperform infrastructure scaling. You’ll leave with... A clear mental model to understand your workload behavior A practical, repeatable framework you can apply immediately to decide when scaling is necessary...and when it’s just unnecessary spend. PUBLIC CONFIRMED Workshops - 2hrs https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/N9BX9T/ Àvila Workshops (Spanish) Guillermo Ruiz PUBLISH RREDSK@@talks.devopsdays.org -RREDSK Multi-Agent Performance Audits: Automatizando Core Web Vitals en tu plataforma interna con MCP es es 20261113T173000 20261113T193000 020000 Multi-Agent Performance Audits: Automatizando Core Web Vitals en tu plataforma interna con MCP **Requisitos técnicos e infraestructura del taller:** Para garantizar una experiencia fluida y sin fricciones durante las 2 horas de taller, la arquitectura está diseñada bajo los siguientes pilares: - Entorno 100% local y contenedorizado: Todo el ecosistema (aplicación de prueba, servidor MCP y automatización del navegador) corre bajo Docker. Los asistentes solo necesitan clonar el repositorio y ejecutar docker-compose up. - Sin dependencias externas ni APIs de pago: Para mitigar problemas de red y costes, el taller utiliza Ollama integrado en Docker con un modelo de código ligero _(qwen2.5-coder o similar)_ precargado en la fase de build. No se requiere ninguna API Key externa. - Entregable "Listo para Producción": El taller es completamente autoguiado mediante un repositorio estructurado paso a paso. Al finalizar, los asistentes se llevarán un entorno real y funcional que podrán replicar y adaptar a las plataformas de desarrollo de sus respectivas empresas al día siguiente. PUBLIC CONFIRMED Workshops - 2hrs https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/RREDSK/ Àvila Workshops (Spanish) Joan Leon PUBLISH EYZDM8@@talks.devopsdays.org -EYZDM8 The Contribution Firehose: Open Source in the Claude era en en 20261114T091500 20261114T095500 004000 The Contribution Firehose: Open Source in the Claude era The 2026 open source ecosystem is facing a paradox of unprecedented scale. On one hand, autonomous AI agents are identifying decades-old vulnerabilities in critical infrastructure and auto-generating patches before threat actors can exploit them. On the other hand, maintainers are being crushed under a DDoS-like wave of security reports and code contributions that cost nothing to generate but demand hours of unpaid human labor to validate. In this keynote, we will explore how global open source communities are adapting to this asymmetric reality. We will look at the rise of the "found on my machine" AI-assisted bug report: complex, unexplainable, and nearly impossible to reproduce. More critically, we will tackle the architectural dilemma: what happens when an AI accurately flags a deep structural flaw, but the refactoring effort is completely out of reach for a volunteer team already strapped for time? While industry initiatives are attempting to pair AI discovery with human security triage to shield maintainers, the human-in-the-loop bottleneck remains a profound threat to project health. Finally, we will zoom out to address an existential shift in software engineering: the fading relevance of traditional licensing. When a developer can simply prompt Claude to build the exact bespoke tooling they need from scratch in seconds, the historical moats of the MIT or GPL licenses begin to evaporate. If the primary abstraction layer has moved from source code to plain-English intent, what does "open source" actually mean today? The primary case study will be the PostgreSQL project. Having worked with and in the project for the better part of 5 years, I have a unique insight into the conversation around adapting for the agents. PUBLIC CONFIRMED Keynote https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/EYZDM8/ Pujades Track (English) Floor Drees PUBLISH WZ9WXY@@talks.devopsdays.org -WZ9WXY Uncovered: The Hard Truth About OpenTelemetry's Vendor Neutrality en en 20261114T100000 20261114T102500 002500 Uncovered: The Hard Truth About OpenTelemetry's Vendor Neutrality Tool choices can shift with reorganizations or mergers. The fact that OpenTelemetry (OTel) is vendor neutral is a superpower…but what exactly does that vendor neutrality actually mean? OTel is designed such that even if you want to swap out OTel’s official SDK for your own, you still can, without breaking anything, as long as you implement the API. The OTel Collector is designed so that you can ingest telemetry data from multiple data sources and multiple formats, and export data using OTLP (OpenTelemetry Protocol) to any OTLP-compatible backend. The ability to mix-and-match OTel-compatible tools, and to pick and choose pieces of OpenTelemetry to adopt, means that build-vs-buy is no longer a binary choice, but a spectrum of choice. End-users are empowered to leverage multiple tools simultaneously, if that best suits their needs. Vendors are empowered to provide ergonomic enhancements on top of open standards. OpenTelemetry is vendor neutral on data ingest. This means that you can point your OTel Collector to two or more OTel-compatible backends. But once the data is ingested… Not all backends may support all 3 “main” OpenTelemetry signals (traces, logs metrics). For example, Jaeger only ingests Traces. Prometheus only ingests metrics. Tools may visualize OTel signals differently Dashboards and dashboard query languages vary across tools. For example, some use PromQL, some use their own proprietary query language, and some use SQL. Some tools optimize on structuring OTel data differently, which can affect pricing when switching to another tool. For example, some vendors encourage sampling. Some discourage sampling. Some encourage enhancing traces with attributes. Some may encourage adding span events to traces. The power of choice means that users can assemble the solution that works for them in pursuit of Observability's true promise: the ability to ask useful questions, get meaningful answers, and enable practitioners to act effectively on what they’ve learned. This talk will unpack OpenTelemetry’s vendor neutrality, and help decision makers and practitioners alike understand what’s truly vendor-neutral and what’s not, helping them select the right tools for their own purposes, and to help them avoid those nasty surprises that come in the form of giant bills, giant learning curves, and interminable migration timelines. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/WZ9WXY/ Pujades Track (English) Josh Lee PUBLISH PN3YNG@@talks.devopsdays.org -PN3YNG Lessons from Building at Massive Scale en en 20261114T103000 20261114T105500 002500 Lessons from Building at Massive Scale Building software at massive scale means designing systems that don’t just work—but continue to work under extreme load, unpredictable traffic spikes, and constant change. In this talk, I’ll share lessons from building and operating distributed systems at companies like Canva and Slack, where platforms handle trillions of requests and support millions of users in real time. We’ll focus on what it actually takes to design for resilience and reliability in high-pressure environments. We’ll cover topics such as: - Designing for high availability and graceful degradation - Handling traffic spikes and load asymmetry without cascading failures - Evolving architectures as scale and complexity grow - Real-world incident response: what breaks, why, and how to recover fast - Applying SRE practices to keep systems reliable at 24x7 global scale This is a practical, experience-driven talk focused on the realities of operating systems at scale—where failure is inevitable, and resilience is designed. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/PN3YNG/ Pujades Track (English) Javier Turégano PUBLISH KHZTA3@@talks.devopsdays.org -KHZTA3 Not on my watch! Catching Web Performance disasters in CI en en 20261114T113000 20261114T115500 002500 Not on my watch! Catching Web Performance disasters in CI In this deep dive, we will explore the 'Not on my watch!' philosophy: building a robust, automated defense against performance regressions. We will cover: - The architectural impact of performance drift in modern web stacks. - How to implement automated performance gates within CI/CD pipelines using GitHub Actions, Lighthouse CI, and Lighthouse CL Server. - Distinguishing between 'lab data' and 'real-user data' when setting up automated thresholds. - Reducing friction in the developer experience: giving feedback early enough to prevent bad code from ever reaching the main branch. Attendees will leave with a clear, actionable framework to set up their own testing infrastructure, minimize production incidents, and foster a culture of performance-first development. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/KHZTA3/ Pujades Track (English) Estela Franco PUBLISH HSBNDT@@talks.devopsdays.org -HSBNDT From backlog to pull request: building an AI-augmented development lifecycle en en 20261114T120000 20261114T122500 002500 From backlog to pull request: building an AI-augmented development lifecycle Many teams lose time in the space between business intent and working software. Backlog items arrive with missing context. Refinement creates clarification loops. Estimation happens before technical implications are fully understood. New joiners depend heavily on senior developers. Developers use AI tools individually, but without a shared process, reusable assets, or measurable delivery impact. This talk presents a real initiative to introduce an AI-augmented development lifecycle into a product team. The goal was not to replace developers or automate delivery end to end. The goal was to improve the flow from backlog item to refined work, from refined work to implementation plan, and from implementation plan to code, while keeping ownership, decision making, and accountability with the team. The first phase introduced an agentic development lifecycle with reusable agents, skills, prompts, templates, and working examples. A refinement agent helps improve clarity, completeness, acceptance criteria, and missing information. A planning agent helps identify dependencies, technical implications, blockers, risks, and estimation inputs. A coding agent uses the implementation plan, the codebase, and team conventions to support code generation and modification. At each step, the human remains in the loop, challenging the output, adjusting it, and validating whether it makes sense. The second phase strengthens the lifecycle with a custom retrieval augmented generation capability for codebase and knowledge access. The idea is to ingest product requirements, software requirements, design specifications, logs, manuals, operational documentation, and source code, then expose this knowledge to both people and agents. With Graph RAG, for example, relationships between code, requirements, business rules, data, and runtime behavior can become easier to explore. A chat interface can make the same knowledge accessible to developers, business analysts, quality engineers, product owners, subject matter experts, and new joiners, regardless of which delivery or development tools they use. The talk will cover the lifecycle design, the human-in-the-loop workflow, the role of reusable prompts and skills, the relationship between agents and retrieval, and the measurement approach. Candidate KPIs include cycle time from ready to done, backlog items completed per sprint, carry over, estimation accuracy, time to find reliable information, onboarding time, repeated questions to senior developers or subject matter experts, token consumption or AI cost per completed backlog item, and avoidable rework. Attendees will leave with a practical model for introducing AI into delivery without treating it as magic, a black box, or a vendor demo. The main takeaway is that AI can improve delivery when it is embedded into team practices, grounded in real context, measured with discipline, and designed to augment people rather than bypass them. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/HSBNDT/ Pujades Track (English) Fabio Scagliola PUBLISH DNTRXX@@talks.devopsdays.org -DNTRXX Flocks vs Pyramids: When self-organization stops working en en 20261114T123000 20261114T125500 002500 Flocks vs Pyramids: When self-organization stops working This talk is aimed at engineers, tech leads, and platform teams dealing with growing coordination complexity between teams. It’s less about a specific framework (like Team Topologies) and more about recognizing the environmental conditions that change how teams collaborate as systems scale. While it touches on themes like autonomy, governance, and architecture, the goal is to help attendees recognize early coordination and communication signals especially if they’re not in formal leadership or architecture roles. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/DNTRXX/ Pujades Track (English) Jos van Schouten PUBLISH 8GWQSV@@talks.devopsdays.org -8GWQSV The 7 Deadly Sins of GitOps en en 20261114T130000 20261114T130500 000500 The 7 Deadly Sins of GitOps We will cover sins such as pride (ignoring drift), envy (Blindly copying other GitOps setups), sloth (ignoring reconciliation errors), and others, as well as practical fixes and patterns that teams can apply to build more resilient and sustainable workflows. PUBLIC CONFIRMED Ignite Talks https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/8GWQSV/ Pujades Track (English) Koray Oksay PUBLISH XFNSAB@@talks.devopsdays.org -XFNSAB Burnout: The Hidden Failure Mode of High-Performing Teams en en 20261114T130500 20261114T131000 000500 Burnout: The Hidden Failure Mode of High-Performing Teams Is your team’s health as observable as your software? High-performing engineering teams are celebrated for their velocity and "whatever it takes" commitment. Yet, these same traits often make them the most vulnerable to a predictable, systemic failure mode: burnout. This isn’t a lack of individual grit; it’s a design flaw. When we architect software for 99.9% uptime but run our teams at 100% capacity, we create a system with zero fault tolerance. In this talk, we move beyond the superficial "resiliency" narrative and analyze team health through the lens of systems engineering. We will explore how to apply SRE principles (like load shedding, redundancy mechanisms, and leading indicators) to the human element of the stack. You’ll learn why "idle capacity" is the secret to low latency, how to make cognitive load visible, and how to build a system where high performance doesn't require a "hero culture" destined for a catastrophic outage. PUBLIC CONFIRMED Ignite Talks https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/XFNSAB/ Pujades Track (English) Diana Nanuti PUBLISH C87ESF@@talks.devopsdays.org -C87ESF From Manual to Autonomous: What Breaks When You Automate Everything en en 20261114T131000 20261114T131500 000500 From Manual to Autonomous: What Breaks When You Automate Everything We set out to automate everything. We never stopped to ask what happens after. As DevOps systems evolve into autonomous, AI‑assisted workflows, failure doesn’t disappear — it changes shape. It becomes quieter, subtler, and far more dangerous. The real risk is no longer the system that breaks, but the one that continues to run… incorrectly and unnoticed. This Ignite talk explores the hidden failure modes of modern automation: where “safe automation” quietly breaks, where assumptions go unchallenged, and where removing humans from the loop shifts — not eliminates — responsibility. Because in a world of autonomous systems, the hardest failures to catch are the ones that look like everything is working. PUBLIC CONFIRMED Ignite Talks https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/C87ESF/ Pujades Track (English) Saikeerthi M PUBLISH WL8Z77@@talks.devopsdays.org -WL8Z77 One Fake Password Is Worth More Than 10,000 Alerts en en 20261114T131500 20261114T132000 000500 One Fake Password Is Worth More Than 10,000 Alerts Security teams have become exceptionally good at generating alerts. Attackers have become exceptionally good at generating noise. As AI-powered tooling lowers the cost of reconnaissance and exploitation, defenders face an uncomfortable reality: more detection does not always mean better detection. This talk argues that high-confidence signals matter more than ever. Using examples from honeypots, canary credentials, decoy files, and deception techniques, we’ll explore why a single interaction with a fake secret can provide more actionable intelligence than thousands of low-confidence alerts. Rather than asking what an attacker could do, deception technologies focus on detecting what an attacker actually did. The result is fewer alerts, higher confidence, and faster response. Key Takeaways: * Why alert volume is becoming a liability. * How deception technologies work in modern environments. * Why AI-assisted attacks increase the value of honeypots. * The difference between exposure-based and intent-based detection. * Practical ideas for implementing low-cost deception techniques. PUBLIC CONFIRMED Ignite Talks https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/WL8Z77/ Pujades Track (English) Ilia Dubovskii PUBLISH VQW9MB@@talks.devopsdays.org -VQW9MB The Carcinisation of AI en en 20261114T132000 20261114T132500 000500 The Carcinisation of AI The jump from Claude to clawdbot, and finally open claw heralded the carcinisation of AI. While crablike adaptations have been successful evolutionary strategy for millions of years and has some early success with AI technology, the humans in the system should be mindful of their crabbiness, especially when it isn't being useful. PUBLIC CONFIRMED Ignite Talks https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/VQW9MB/ Pujades Track (English) ecology_chris PUBLISH Z9L8GU@@talks.devopsdays.org -Z9L8GU Open Space 1 en en 20261114T153000 20261114T160000 003000 Open Space 1 Open Space 1 PUBLIC CONFIRMED Open Space https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/Z9L8GU/ Pujades Track (English) PUBLISH YW8EKY@@talks.devopsdays.org -YW8EKY Open Space 2 en en 20261114T163000 20261114T170000 003000 Open Space 2 Open Space 2 PUBLIC CONFIRMED Open Space https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/YW8EKY/ Pujades Track (English) PUBLISH QRKUEB@@talks.devopsdays.org -QRKUEB The Hardest Platform Problems Weren't Technical en en 20261114T100000 20261114T102500 002500 The Hardest Platform Problems Weren't Technical Platform Engineering is often presented as a problem of tooling. The discussion usually revolves around self-service capabilities, golden paths, developer portals, automation and developer experience. Those capabilities are important and we invest in many of them ourselves, but over the last few years I have come to believe that they are rarely the limiting factor. In my experience, platform teams do not suffer from a lack of ideas. The opposite is usually true. Reliability improvements, developer experience initiatives, compliance requirements, cost optimization, technical debt reduction and new platform capabilities can all have a reasonable business case behind them. The difficulty is that they all compete for the same limited capacity. At Cabify we have been trying to evolve from infrastructure as a collection of services towards Platform Engineering with a stronger product mindset. That journey has been useful because it helped make some of these tensions more visible. We found ourselves spending less time discussing how to implement things and more time discussing why we should implement them, what value they would create and what other work would be displaced as a consequence. As the number of opportunities grew, we started experimenting with different approaches to prioritization. We introduced a single backlog for Platform, moved discussions upstream, explored Internal Product Advocates (IPAs) and tried to make value and expected outcomes more explicit. None of these approaches magically solved the problem, but they helped create better conversations around trade-offs and priorities. In this talk I will share examples from infrastructure and platform work, including operational improvements, reliability initiatives, governance requirements and developer experience investments. The goal is not to present a prioritization framework or claim that we found the answer. Rather, I want to discuss a challenge that I suspect many platform teams face: how do you make decisions when most of the available options are worthwhile and there is never enough capacity to do all of them? My hope is that attendees leave with a different perspective on Platform Engineering. Beyond the technology, the tooling and the automation, there is an ongoing exercise in deciding where to invest limited capacity. Those decisions shape the platform at least as much as any architectural choice, yet they receive far less attention in our industry discussions. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/QRKUEB/ Llacuna Track (English) Abel Navarro PUBLISH QKVH33@@talks.devopsdays.org -QKVH33 DevOps in Space: Lessons from Low Earth Orbit en en 20261114T103000 20261114T105500 002500 DevOps in Space: Lessons from Low Earth Orbit The Infrastructure & Software Engineering team at Sateliot works at the intersection of aerospace, telecommunications and software. We build modern distributed systems and then send them to an extremely harsh environment, where normal expectations -low latency, immediate feedback, up-to-date systems- simply don’t apply. In this talk I’ll give a high-level tour of the challenges that make space and telco different: intermittent and low-bandwidth connectivity, long feedback loops, constrained devices, strong regulation and the cultural gaps between very different fields of engineering. I'll also share the ways we have found within those constraints to apply some of the DevOps processes and tools that have become industry standards. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/QKVH33/ Llacuna Track (English) David Jacovkis PUBLISH ZTQYMS@@talks.devopsdays.org -ZTQYMS Access Without Chaos: SQL Guardrails When Everyone's a Builder en en 20261114T113000 20261114T115500 002500 Access Without Chaos: SQL Guardrails When Everyone's a Builder Mews had grown organically for over a decade through product-market fit and scale stages without any effective SQL access governance. So it was that we came to this problem with a chaotic mess of access and permissions across our databases. We didn't exhibit least privilege, there was no logging of access, and no shared standards. In many cases, it was guesswork on who had access to each database. That was the problem worth solving. The solution came together in layers. We built a centralised Terraform repository as the single source of truth for SQL access configuration. We introduced a PIM self-service model where engineering managers and staff engineers manage their own teams' database memberships - no Platform team involvement needed at runtime. 400+ engineers now operate within a zero-standing-access model: time-limited, justified, fully auditable. Connection strings and local users gave way to managed identity authentication throughout the fleet. Three cloud-native policies now enforce compliance at the control plane level - resource locks preventing accidental deletion, identity administrator governance, and audit logging routed to a centralised store. These aren't conventions or recommendations. Teams cannot remove or override them. The shared Terraform SQL module provides opinionated defaults that land compliant from day one, but it's the policy layer that makes compliance permanent. New services can't drift even if they try. In 2026, with AI tools enabling everyone - engineers, analysts, and non-technical contributors - to spin up infrastructure and access production data, the blast radius of poor access hygiene has never been wider. Carlos will walk you through the full journey - from the first Terraform commit to production-enforced cloud policies - including what broke along the way, what we over-engineered the first time, and the organisational side of handing access ownership back to product teams. **You'll leave with:** - A blueprint for self-service database access using just-in-time privilege elevation and identity provider groups - A clear picture of how cloud-native policy enforcement makes compliance immutable - not just recommended - Honest lessons from the rollout: what broke, what surprised us, and the org change management side of removing the Platform team from the critical path - A framework for thinking about access guardrails in the AI era - where anyone can become a data consumer, and the blast radius of poor governance has never been wider This talk is for platform engineers, DevSecOps practitioners, and anyone who has ever found themselves as the bottleneck on access governance. _No access requests were harmed in the making of this talk._ PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/ZTQYMS/ Llacuna Track (English) Carlos Augusto Pega PUBLISH WNP8ZE@@talks.devopsdays.org -WNP8ZE 🧠 The developer’s brain: Cognitive science for better code en en 20261114T120000 20261114T122500 002500 🧠 The developer’s brain: Cognitive science for better code If you’re a developer, your most important tool is not your editor or your setup but your brain. And yet, most people don't know much about how it actually works. This talk is a practical introduction to the cognitive side of software engineering. We’ll explore why coding is so mentally demanding and how understanding a few key mechanisms of the brain can significantly improve the way we work. Through real-world examples and relatable scenarios, we’ll cover four essential concepts: 1. The brain’s background processor - Zooming in vs zooming out - How to trigger the "a-ha" moments 2. Neuroplasticity - What happens in the brain when we learn new technologies - What's the most optimal strategy to learning 3. Cognitive load - Why complex code quickly becomes overwhelming - Techniques to manage mental load through decomposition and externalization 4. Biases - The brain as a pattern recognition machine - A bias-resistent approach to debugging and problem-solving We’ll finish by combining these ideas into a practical cognitive toolkit you can apply immediately in your day-to-day work. This isn’t a theory-heavy talk, but rather some practical guidelines to building a cognitive toolkit you can apply immediately in your day-to-day work so you can write better code, make better decisions, and work more sustainably. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/WNP8ZE/ Llacuna Track (English) Marina Altés PUBLISH WKDJVB@@talks.devopsdays.org -WKDJVB Extra Crispy: Your Incident Process Is Burning Out Your Engineers en en 20261114T123000 20261114T125500 002500 Extra Crispy: Your Incident Process Is Burning Out Your Engineers At 2:13 AM, nobody cares whether your incident process is technically compliant. They care whether people know what to do, whether communication is clear, and whether the same exhausted engineer has become the single point of failure for your entire company. For years, many organizations treated ITIL and DevOps as enemies: one focused on control, the other on speed. But after working with engineering and operations teams across multiple environments, I discovered the real problem wasn’t process, it was designing systems that ignored the humans inside them. In this talk, I’ll share real-world lessons from teams struggling with alert fatigue, endless escalations, hero culture, and burnout disguised as “ownership.” We’ll explore how modern ITIL concepts like continual improvement, incident management, and value streams can actually support DevOps practices when implemented with empathy instead of bureaucracy. ## Attendees will learn: - Why “hero culture” silently destroys operational resilience - How poorly designed escalation paths increase burnout - What psychologically safe incident response actually looks like - Where ITIL still provides value in modern DevOps organizations - Practical ways to reduce operational stress without sacrificing reliability PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/WKDJVB/ Llacuna Track (English) Chris Ward PUBLISH 8GQAKJ@@talks.devopsdays.org -8GQAKJ Open Space 3 en en 20261114T153000 20261114T160000 003000 Open Space 3 Open Space 3 PUBLIC CONFIRMED Open Space https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/8GQAKJ/ Llacuna Track (English) PUBLISH BXVXCB@@talks.devopsdays.org -BXVXCB Open Space 4 en en 20261114T163000 20261114T170000 003000 Open Space 4 Open Space 4 PUBLIC CONFIRMED Open Space https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/BXVXCB/ Llacuna Track (English) PUBLISH UMTHCM@@talks.devopsdays.org -UMTHCM GenAI in a box: Construyendo una Plataforma de IA en Kubernetes es es 20261114T100000 20261114T102500 002500 GenAI in a box: Construyendo una Plataforma de IA en Kubernetes Diseñada para SREs e Ingenieros de Plataforma, esta sesión busca abordar la realidad de operar IA en producción dentro de una plataforma automatizada abordando: - FinOps y Soberanía: Alternativas viables a las APIs SaaS para cargas de trabajo sensibles y agentes autónomos de alto consumo. - Arquitectura en K8s: El LLM (Gemma) como carga stateless, Model Context Protocol (MCP) para integración segura, y PostgreSQL como motor stateful. - Operando Postgres + pgvector en K8s: Cómo gestionar los picos de recursos (CPU/RAM) que introducen las búsquedas vectoriales para evitar la inestabilidad del clúster, y cómo automatizar el almacenamiento persistente. - Platform Engineering (GitOps): Cómo empaquetar toda esta complejidad bajo un enfoque declarativo para ofrecer entornos de IA aislados en formato "autoservicio" a los desarrolladores en cuestión de minutos. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/UMTHCM/ Rambla Track (Spanish) Miguel Toscano PUBLISH CSQ3R7@@talks.devopsdays.org -CSQ3R7 Código y Cerebro: Salud y Rendimiento en Entornos Tecnológicos de Alta Demanda Cognitiva es es 20261114T103000 20261114T105500 002500 Código y Cerebro: Salud y Rendimiento en Entornos Tecnológicos de Alta Demanda Cognitiva Los entornos tecnológicos, especialmente en contextos de DevOps y comunidades open source, requieren cambios constantes de contexto, multitarea y exposición continua a altos niveles de carga cognitiva. Esto impacta directamente la atención, la toma de decisiones, el rendimiento y la salud mental. La charla se basa en experiencias reales con profesionales y comunidades tech, donde se observan patrones recurrentes: fatiga mental, sobrecarga, dificultad de concentración, inestabilidad en el rendimiento y aumento de vulnerabilidades en el trabajo diario. Desde la neurociencia aplicada, se analiza cómo la carga cognitiva, las interrupciones, la presión por entregas y el exceso de información afectan las funciones ejecutivas y el comportamiento. Más allá de la teoría, la sesión propone aplicación práctica. Se presentan estrategias utilizadas en contextos reales, como técnicas de regulación cognitiva, organización del trabajo y construcción de entornos más sostenibles. También se discute cómo la cultura y la dinámica de los equipos pueden amplificar o reducir estos efectos. La sesión se realizará en inglés y español, promoviendo el intercambio entre participantes. Se invita a compartir experiencias y construir colectivamente estrategias para el cuidado y el rendimiento sostenible. Esta charla fue presentada en KubeCon Europe 2026 (Ámsterdam) y en eventos DevOps en Brasil (Salvador y Florianópolis, 2025), formando parte de una trayectoria de participación en casi 30 eventos tecnológicos desde 2022. Público objetivo: profesionales de tecnología, comunidades open source, líderes técnicos y equipos de alto rendimiento en entornos de alta demanda cognitiva. Formato: charla interactiva con espacio para la aplicación de ejercicios de regulación neurocognitiva. Ejes principales: Carga cognitiva y funcionamiento cerebral Impactos en la salud y el rendimiento en tecnología Estrategias prácticas de regulación neurocognitiva y organización del trabajo Construcción de entornos sostenibles e inclusivos Diferencial: integración entre neurociencia aplicada, experiencia práctica en comunidades tech y enfoque orientado a la implementación real. Adaptación: la charla amplía la teoría hacia actividades prácticas de salud. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/CSQ3R7/ Rambla Track (Spanish) Dri Cardoso João Brito PUBLISH KA7ZWQ@@talks.devopsdays.org -KA7ZWQ De los dashboards a las decisiones: integración de IA en operaciones DevOps a gran escala es es 20261114T113000 20261114T115500 002500 De los dashboards a las decisiones: integración de IA en operaciones DevOps a gran escala Modern DevOps teams rarely suffer from a lack of data. **The real problem is fragmentation.** During an incident, engineers may need to move between **Kubernetes, AWS, Argo CD, observability platforms, CI/CD pipelines, repositories, edge services, and internal documentation**. Each system provides part of the answer, but reconstructing the full operational story still requires time, experience, and repeated manual investigation. This talk presents our experience introducing **AI-assisted operational workflows** into a global e-commerce environment running distributed services across multiple platforms and environments. Rather than using AI only to generate commands or summarize logs, we explored how it could act as an **investigation layer across existing DevOps systems**: collecting context, correlating events, proposing hypotheses, and helping engineers move from raw telemetry to an informed operational decision. ### What the session will cover - The operational challenges of a **large-scale e-commerce platform** with multiple services, environments, deployment pipelines, and observability tools. - How operational context can be gathered from **Kubernetes, AWS, Argo CD, monitoring platforms, repositories, and deployment history**. - How an AI-assisted investigation loop can **plan actions, call read-only tools, evaluate observations, and refine its hypothesis**. - Why **deterministic workflows and AI agents** solve different categories of operational problems. - How we introduced **access controls, environment boundaries, auditability, stopping conditions, and human approval points**. - What failed during implementation, including **incomplete context, misleading correlations, excessive tool calls, and confident but incorrect conclusions**. - How tracing and evaluation helped us understand whether the system was actually improving incident investigation. - Which tasks proved useful for AI assistance and which remained safer and more reliable as traditional automation. A central theme of the talk is that **connecting a language model to operational tools is the easy part**. The difficult part is designing a system that knows: 1. What it is allowed to inspect. 2. How to validate what it finds. 3. When to stop. 4. When to defer to an engineer. The session will include a **practical architecture**, an example incident-investigation flow, the controls required for enterprise use, and the lessons we learned while applying these ideas in a real production environment. ### Key takeaways Attendees will leave with: - A realistic framework for deciding where AI can add value in DevOps. - A clearer understanding of where deterministic automation remains the better choice. - Practical guidance on the technical safeguards required before introducing agents into operational workflows. - Lessons learned from applying AI-assisted investigation in a real enterprise environment. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/KA7ZWQ/ Rambla Track (Spanish) JOSE JULIAN ARIZA V. PUBLISH VWTFSC@@talks.devopsdays.org -VWTFSC Entitlements as Code: gobernanza de accesos para la era cloud-native es es 20261114T120000 20261114T122500 002500 Entitlements as Code: gobernanza de accesos para la era cloud-native Las organizaciones cloud-native han adoptado prácticas declarativas para infraestructura, despliegues y políticas, pero la gestión de accesos sigue dependiendo a menudo de tickets, procesos manuales y conocimiento tribal. Esta charla presenta Entitlements as Code: una forma de gestionar usuarios, grupos y membresías como estado deseado en Git, usando Entra ID como hub de identidad y un reconciler como Crossplane para alinear continuamente la definición declarativa con la realidad. A través de una evolución práctica, veremos cómo pasar de un repositorio versionado a capas de autoservicio mediante scripts, portales, formularios o asistentes de IA, sin romper el flujo de revisión, aprobación y auditoría. Los asistentes entenderán cómo reducir cuellos de botella, eliminar dependencias de personas concretas y construir un modelo de gobernanza de accesos más trazable, escalable y sostenible. PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/VWTFSC/ Rambla Track (Spanish) Edu Borrego PUBLISH XY3T7L@@talks.devopsdays.org -XY3T7L Kafka a escala: cómo pasar de un mapa de metro a una arquitectura simplificada es es 20261114T123000 20261114T125500 002500 Kafka a escala: cómo pasar de un mapa de metro a una arquitectura simplificada Esta charla presenta la evolución real de una plataforma de procesamiento de eventos que fue creciendo durante años hasta que la complejidad operativa se convirtió en uno de los principales problemas a resolver. La arquitectura original estaba basada en Kafka y una base de datos MySQL compartida donde persistían datos los distintos consumidores. A medida que aumentó el volumen de eventos, la base de datos central empezó a convertirse en un cuello de botella, así que se optó por distribuir la carga creando bases de datos independientes para distintos consumidores. Esta solución funcionó durante bastante tiempo y permitió continuar escalando la plataforma y absorver más volumen de eventos. Sin embargo, cada nueva necesidad acababa incorporando nuevas bases de datos, nuevos procesos de agregación y nuevos puntos de integración. Lo que inicialmente era una optimización razonable terminó derivando en un ecosistema formado por más de 20 bases de datos y múltiples servicios agregadores encargados persistir los datos en la base de datos principal. Con el paso del tiempo empezaron a aparecer problemas que probablemente resulten familiares para muchos equipos, por ejemplo los costes operativos derivados de esta infraestructura, la dificultad en la gestión o las dependencias cruzadas entre sistemas, e incluso limitaciones que impactaban gravemente en negocio. Para resolver esta situación decidimos replantear la arquitectura y construir una plataforma analítica centralizada basada en Kafka, Kafka Connect y ClickHouse Durante la sesión hablaremos de - cómo evolucionó la arquitectura original - qué problemas intenta resolver cada una de las decisiones tomadas - en qué momento la complejidad pasó a ser más costosa que los beneficios obtenidos - cómo diseñamos y ejecutamos la migración - los retos encontrados durante el proceso - los resultados obtenidos tanto en performance como en gestión - los retos que aún nos quedan pendientes Sin embargo, una parte importante de esta historia no está en los diagramas. También hablaremos del factor humano detrás de la transformación: cómo se construye un equipo capaz de abordar cambios de gran escala, la importancia de la formación continua, la transferencia de conocimiento y la creación de una cultura de colaboración. Aprendimos que la simplificación tecnológica no sucede de un día para otro y que la cohesión del equipo es tan importante como la arquitectura cuando se trata de hacer evolucionar una plataforma crítica Porque al final, las tecnologías cambian pero son los equipos los que permiten que sistemas complejos sigan evolucionando de forma sostenible PUBLIC CONFIRMED Talk https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/XY3T7L/ Rambla Track (Spanish) Esther Yébenes PUBLISH 9FCUAQ@@talks.devopsdays.org -9FCUAQ Open Space 5 es es 20261114T153000 20261114T160000 003000 Open Space 5 Open Space 5 PUBLIC CONFIRMED Open Space https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/9FCUAQ/ Rambla Track (Spanish) PUBLISH SYYKKP@@talks.devopsdays.org -SYYKKP Open Space 6 es es 20261114T163000 20261114T170000 003000 Open Space 6 Open Space 6 PUBLIC CONFIRMED Open Space https://talks.devopsdays.org/devopsdays-barcelona-2026/talk/SYYKKP/ Rambla Track (Spanish)