Articles from Source: Red-Hat-Developer-Blog

What's new in Red Hat Ansible Automation Platform 2.7

2026-06-10 14:05
🚀 Red Hat Ansible Automation Platform 2.7 is here! This release focuses on enhancing automation for IT teams at enterprise scale. Key updates include a new visual execution environment builder and centralized content catalog to streamline processes. The intelligent assistant now supports bring-your-own-knowledge for tailored guidance. Additionally, the MCP server allows AI agents to manage automation through natural language commands, improving efficiency. Explore the new features and start...
Source: Red Hat Developer Blog
Catherine Choi

What's new in Red Hat Ansible Automation Platform 2.7

2026-06-10 14:05
🚀 Red Hat Ansible Automation Platform 2.7 is now available, enhancing efficiency and intelligence for IT teams. Key updates include: 🔹 A visual execution environment builder for streamlined automation setup. 🔹 A centralized content catalog to easily access trusted collections. 🔹 Ansible development workspaces provide a consistent, browser-based environment. Additionally, the platform introduces an MCP server for AI-driven automation queries and a new intelligent assistant that supports custom...
Source: Red Hat Developer Blog
Catherine Choi

Building and running Bazel applications on AutoSD: Toolchains, containers, and recommended practices

2026-06-10 13:09
🚀 Bazel is an open-source build system that streamlines software builds and tests across multiple languages and platforms. This article discusses methods for building Bazel applications on Automotive Stream Distribution (AutoSD). Three approaches are highlighted: 1️⃣ **Native AutoSD GCC Toolchain**: Direct integration with AutoSD for minimal abstraction. Best for exclusive AutoSD targets. 2️⃣ **S-Core Abstracted Toolchain**: Supports multiple platforms, ideal for CI/CD environments, reducing...
Source: Red Hat Developer Blog
Bilal Elmoussaoui

Bring your own evaluation framework to EvalHub

2026-06-09 07:01
🚀 EvalHub now supports a "bring-your-own-framework" (BYOF) approach, allowing teams to integrate custom evaluation frameworks. This enables organizations to leverage proprietary or academic evaluation harnesses not included in the default provider set. By implementing a simple Python method, users can package their framework, enabling features like experiment tracking and OCI artifact persistence. Learn more about building your custom adapter and the integration process. #EvalHub #AI...
Source: Red Hat Developer Blog
William Caban Babilonia, Rui Vieira, Matteo Mortari

Integrate OpenShift AI and PG Airman MCP Server

2026-06-09 07:01
🚀 Discover how agentic AI is transforming data access in enterprises! This article introduces the integration of Red Hat OpenShift AI and EnterpriseDB’s PG Airman MCP server, addressing the challenges non-technical staff face with SQL. It highlights a natural language interface that simplifies data queries and ensures compliance with data governance. Stay tuned for more insights in this four-part series! #DataGovernance #AI #PostgreSQL #OpenShift #TechInnovation
Source: Red Hat Developer Blog
Peter Samouelian

Build a local voice agent with Red Hat OpenShift AI

2026-06-08 07:01
🚀 Building a local voice agent with Red Hat OpenShift AI can be complex but rewarding. This article provides insights on creating a pizza shop voice agent, focusing on architecture, model selection, and implementation challenges. Key steps include using a voice sandwich architecture, selecting models from the Red Hat OpenShift AI catalog, and ensuring quick execution for natural conversation flow. Explore the full process and learn about performance metrics, agent collaboration, and the...
Source: Red Hat Developer Blog
Mike Hepburn

Gang autoscaling on OpenShift with Kueue and ProvisionRequest

2026-06-08 07:01
Exploring gang autoscaling on OpenShift is crucial for high-performance workloads like AI/ML training. Traditional Kubernetes scheduling can lead to resource waste when pods can't start simultaneously due to capacity limits. The combination of Red Hat's Kueue and the ProvisionRequest API addresses this issue by coordinating resource availability before scheduling. This ensures that all required pods start together, optimizing resource use. For a deep dive into the setup and benefits, check...
Source: Red Hat Developer Blog
Kevin Hannon, Michael McCune

Installing Red Hat Enterprise Linux 10 from a bootc image with bootc

2026-06-05 03:01
🔧 Interested in managing Red Hat Enterprise Linux (RHEL) more efficiently? The new image mode deployment option allows installation from a bootc image, enhancing consistency and enabling atomic updates. 🖥️ With RHEL 10, you can easily deploy this system using Anaconda and the new bootc kickstart command. This method streamlines the installation process by managing core tasks while ensuring system updates can be performed later. 📥 For more details on how to get started, check the full article...
Source: Red Hat Developer Blog
Jiří Kortus

Why your database benchmarking data is probably wrong (and how I fixed mine)

2026-06-05 03:01
🔍 Struggling with database benchmarking? You're not alone. An article outlines common pitfalls faced when testing AWS RDS PostgreSQL performance. One key issue is the load generator acting as a bottleneck, impacting throughput. Upgrading the client instance can help eliminate this limitation. Another factor is ensuring the test is CPU-bound rather than disk-bound by adjusting parameters. Additionally, increasing the max_wal_size can prevent performance dips during testing. For reliable...
Source: Red Hat Developer Blog
Krishna Magar

Type what you want to break: AI-assisted chaos engineering with Krkn

2026-06-04 07:16
Unlock the power of chaos engineering with Krkn! 🚀 Krkn now supports over 20 scenario types for Kubernetes, including pod disruptions and network chaos. However, translating your testing intent into precise CLI syntax can be challenging. A new solution simplifies this: a natural language interface that generates validated Krkn commands from plain English. Just describe the failure you want to simulate, and the tool handles the syntax for you. This innovation enhances accessibility, allowing...
Source: Red Hat Developer Blog
Darshan Jain

Understanding evaluation collections in EvalHub

2026-06-04 07:16
Discover the importance of evaluation collections in AI with EvalHub! This article discusses common pitfalls in AI evaluation, highlighting the need for precise metrics tailored to your deployment context. It introduces evaluation-driven development and how to create personalized collections that meet your specific criteria. The Leaderboard v2 collection serves as a practical example, showcasing how to define benchmarks, weights, and thresholds effectively. Explore how to build a robust...
Source: Red Hat Developer Blog
William Caban Babilonia, Julian Payne, Marius Ion Danciu

An overview of confidential containers on OpenShift bare metal

2026-06-04 07:16
Discover how Confidential Containers leverage Trusted Execution Environments (TEEs) on OpenShift bare metal for enhanced workload isolation. At the core are confidential virtual machines (CVMs) that utilize Kata Containers for running Kubernetes pods, ensuring strong security through hardware isolation. 🔒 Remote attestation verifies the integrity of CVMs, ensuring sensitive materials are securely handled. This architecture supports a zero-trust model, enhancing confidentiality and integrity...
Source: Red Hat Developer Blog
Pradipta Banerjee, Leonardo Milleri, Emanuele Giuseppe Esposito, Pei Zhang

iSCSI vs. NVMe/TCP: The ultimate storage showdown for Red Hat OpenShift Virtualization

2026-06-04 07:16
🔍 In the latest article, we explore the comparison between iSCSI and NVMe/TCP storage protocols in Red Hat OpenShift Virtualization. Both protocols have distinct advantages. iSCSI has been a reliable choice for years but may struggle with modern SSDs due to its single-queue architecture. In contrast, NVMe/TCP is optimized for high-performance flash storage, offering lower latency and higher IOPS. Testing shows NVMe/TCP significantly outperforms iSCSI in VM provisioning and raw disk I/O,...
Source: Red Hat Developer Blog
Sonali Badal

Speculators v0.5.0: DFlash support and online training

2026-06-04 07:16
🚀 Exciting news! The v0.5.0 release introduces major upgrades for speculative decoding model training. Key features include DFlash algorithm support, which enables single-pass draft token generation, and enhanced online training capabilities. The integration with vLLM’s hidden states extraction system streamlines both online and offline training. Updated documentation provides clear guidance for users. Explore the benefits of the DFlash algorithm and its performance in various tasks....
Source: Red Hat Developer Blog
Helen Zhao, Fynn Schmitt-Ulms, Dipika Sikka

Intelligent inference scheduling with llm-d on Red Hat AI

2026-06-04 03:01
Discover how intelligent inference scheduling with llm-d enhances AI performance on Red Hat platforms. The article explores the benefits of optimizing scheduling processes to improve efficiency and resource management in AI applications. Learn how Red Hat AI is leveraging these advancements for better outcomes. #RedHatAI #ArtificialIntelligence #TechInnovation #InferenceScheduling #llmD 🤖📈💡
Source: Red Hat Developer Blog
Edoardo Vacchi

Build modular AI pipelines with OpenShift AI and reusable components

2026-06-03 07:31
🚀 Red Hat OpenShift AI enables teams to build modular AI pipelines using reusable components, streamlining the development process. These standardized building blocks help handle tasks like data preprocessing, model training, and deployment, saving time and reducing duplicated efforts. By leveraging a shared component library, teams can enhance collaboration and ensure consistency across projects. Explore the benefits of composable AI workflows today! 🌐 #RedHat #OpenShiftAI #AIPipelines...
Source: Red Hat Developer Blog
Ana Biazetti, Nelesh Singla, Matt Prahl

UBI 9 and 10 builders on Paketo Buildpacks with multi-arch support

2026-06-02 07:01
🚀 Exciting updates from Paketo Buildpacks! UBI 9 and UBI 10 builders are now available, offering multi-architecture support for builds. These builders currently support Node.js, with Java support on the way. Base images are published on Dockerhub, making it easy to get started. Multi-arch options include arm64 and amd64, allowing for flexible application builds. For more details, check the release notes and try using the pack CLI! #PaketoBuildpacks #UBI #MultiArch #NodeJS #DevOps
Source: Red Hat Developer Blog
Costas Papastathis

Deploy Hermes Agent on OpenShift AI with vLLM model serving

2026-06-02 07:01
🚀 Exciting advancements in AI deployment! The article outlines how to deploy the Hermes Agent on Red Hat OpenShift AI using GPU-accelerated vLLM model serving. This innovative agent retains user context across sessions, creating a continuous learning experience. Key features include: - Multi-platform capabilities (Telegram, Discord, Slack) - Self-improving skills from multi-step tasks - Seamless integration with OpenShift's production-ready AI infrastructure This deployment is ideal for...
Source: Red Hat Developer Blog
Gerald Trotman

Evaluation-driven development with EvalHub

2026-06-02 07:01
🚀 Discover the future of AI development with Evaluation-Driven Development (EDD) using EvalHub! EDD transforms traditional test-driven development by focusing on measurable performance gaps instead of simple pass/fail outcomes. 🔍 Key Steps in EDD: 1️⃣ Define clear evaluation criteria before coding. 2️⃣ Measure quality with gradient scores for deeper insights. 3️⃣ Iterate based on data to optimize performance. EvalHub streamlines this process, ensuring effective AI outcomes through transparent...
Source: Red Hat Developer Blog
William Caban Babilonia, Matteo Mortari

Improve vLLM Semantic Router accuracy with fine-tuning

2026-06-02 07:01
🚀 The vLLM Semantic Router enhances model efficiency by routing requests to the appropriate models based on complexity. However, a recent study found that the pretrained model had an 80% accuracy rate, leading to a 20% misrouting rate. This highlights a critical need for improved accuracy in enterprise deployments. To address this, a fine-tuning pipeline was established on OpenShift AI, significantly boosting routing accuracy from 80% to 98.5%. This adjustment ensures that models handle...
Source: Red Hat Developer Blog
Christopher Nuland

Red Hat build of Cryostat 4.2: Enhanced Java monitoring for OpenShift

2026-06-02 07:01
🚀 The Red Hat build of Cryostat 4.2 is now generally available, enhancing Java monitoring on OpenShift. 📊 Key features include SQL query support for JDK Flight Recorder data, allowing for in-console analysis without downloading files. 🔍 The update also introduces async-profiler integration for better stack trace capture and smart triggers for dynamic recording management. 🔒 New observability features like audit logging and improved thread dump analysis enhance security and system tracking....
Source: Red Hat Developer Blog
Syed M Shaaf, Chris Mah

Protect your Kubernetes Operator from OOMKill

2026-06-01 07:01
🛡️ Protecting your Kubernetes Operator from OOMKill is crucial. Kubernetes operators, which manage applications automatically, have a vulnerability linked to unfiltered informer caches. This can lead to memory exhaustion and crash your operator, exposing it to potential denial-of-service attacks. To mitigate this, ensure your cache is filtered by labels and implement best practices during updates. Learn more about safeguarding your cluster! #Kubernetes #DevOps #CloudComputing #Security...
Source: Red Hat Developer Blog
Rishabh Singh, Ugo Giordano

Owning the system clock: Good enough?

2026-06-01 03:01
Accurate timing is crucial across various industries, as applications rely on the system clock to reflect real-world time. The challenge lies in achieving consistent accuracy everywhere. ⏰ Most systems use Network Time Protocol (NTP) for millisecond accuracy, while Precision Time Protocol (PTP) can reach up to 100 nanoseconds. Global Navigation Satellite System (GNSS) is another option, but it faces risks like jamming. 🌍 To ensure reliability, a solution combines GNSS as the primary source...
Source: Red Hat Developer Blog
Joseph Richard

What's new in OpenShift Container Platform system management

2026-05-29 07:01
🔍 Red Hat OpenShift Container Platform introduces key updates in system management! Starting with version 4.21, new clusters will automatically allocate system-reserved resources based on node size, addressing past memory and CPU competition issues. Additionally, version 4.22 introduces CPU limit enforcement for system daemons, enhancing stability. These changes aim to improve node performance while maintaining compatibility with existing setups. #OpenShift #RedHat #Kubernetes #CloudComputing...
Source: Red Hat Developer Blog
Neeraj Krishna Gopalakrishna

Claude as your performance analysis partner

2026-05-29 03:01
Unlock the potential of performance analysis with Claude! 🚀 This article explores how Claude simplifies the challenging task of analyzing large CPU profiles and traces, particularly with the Go Green Tea garbage collector. It highlights how Claude identifies bottlenecks and suggests optimizations effectively. Key aspects include analyzing CPU profiles using Go's pprof tool and optimizing atomic operations for better performance. Claude also aids in recognizing patterns in trace files to...
Source: Red Hat Developer Blog
Archana Ravindar

LogAn: Large-scale log analysis with small language models

2026-05-28 07:16
🚀 Introducing LogAn: a new approach to log analysis that addresses the limitations of Large Language Models (LLMs). Traditional LLMs struggle with the vast volume of log data, often processing mostly routine messages rather than critical errors. LogAn offers a solution by utilizing a template mining algorithm called Drain, which compresses logs into unique templates for efficient analysis. Developed by IBM Research and open-source, LogAn combines log templatization and semantic analysis to...
Source: Red Hat Developer Blog
Rahul Shetty, Aman Vishwakarma

stalld’s BPF Backend: Breaking Free from debugfs

2026-05-28 03:01
🚀 Exciting updates for stalld! The new BPF-based queue_track backend enhances task starvation detection in Linux environments. By shifting from a poll-based method to an event-driven model, stalld improves efficiency and reliability. This change eliminates reliance on debugfs, ensuring better performance and compatibility across kernel versions. Learn more about how this evolution supports real-time workloads! 🔧💻 #Linux #BPF #stalld #TechUpdate #OpenSource
Source: Red Hat Developer Blog
Clark Williams, Wander Lairson Costa

Running AI inference on Rebellions ATOM NPU with Red Hat AI

2026-05-27 07:16
🚀 Exciting news for enterprises scaling AI! Red Hat and Rebellions have launched a joint solution integrating Rebellions' ATOM NPUs with Red Hat OpenShift AI. This collaboration enables efficient AI inference with low latency and high throughput. 🔍 The ATOM NPU, designed for AI workloads, offers significant energy savings compared to traditional GPUs. This allows organizations to optimize costs while maintaining performance. 📈 The solution includes a comprehensive architecture for deploying...
Source: Red Hat Developer Blog
Erwan Gallen, Chris Procter, Liming Tsai

How we built integration testing for fast-moving AI backend

2026-05-27 07:16
🚀 Keeping up with rapidly changing APIs can be a challenge. At Red Hat OpenShift AI, we faced this issue with Llama Stack, where mocked unit tests failed to reflect real-time changes. To solve this, we integrated a real Llama Stack server into our testing. By using its record-replay functionality, we avoided costly LLM calls while ensuring reliability in our tests. Now, our daily workflow includes a Slack sentinel that alerts us about compatibility, giving us early warnings on potential...
Source: Red Hat Developer Blog
Avik Kundu

Testing infrastructure red teaming with abliterated models

2026-05-26 07:01
🔍 Testing the security of agent workloads on Red Hat OpenShift has revealed critical insights. The study deployed OpenClaw and utilized custom probes across various attack categories. Five models were tested, with abliterated models showing 100% cooperation on adversarial prompts, highlighting the importance of infrastructure as a final security measure. Key findings include: - Tier 0 (no controls) saw significant credential exfiltration. - Adding an SSH sandbox (Tier 1) eliminated sensitive...
Source: Red Hat Developer Blog
Roy Belio

Build an enterprise RAG system with OGX

2026-05-26 07:01
Transforming naive RAG systems into effective solutions is possible with OGX (Open GenAI Stack). Key strategies include: 🔍 **Metadata Filtering**: Enhances accuracy by narrowing search results based on user context. 🔄 **Hybrid Retrieval**: Combines vector and keyword searches for precise results. 📊 **Neural Reranking**: Uses advanced models to improve relevance in responses. Explore how these techniques enhance enterprise RAG applications! #RAG #OpenSource #DataRetrieval #AI #MachineLearning
Source: Red Hat Developer Blog
Abdelhamid Soliman

Solutions for SELinux MCS challenges with GitLab runners

2026-05-26 07:01
🚀 SELinux Multi-Category Security (MCS) poses challenges for GitLab runners by restricting access between containers sharing volumes. GNOME addresses this issue with fixed MCS labels but compromises isolation. The article explores potential solutions, including microVM isolation using Cloud Hypervisor and Firecracker. Discover how these technologies may enhance security in CI environments. 🔍💻 #SELinux #GitLab #MicroVM #CloudHypervisor #DevOps
Source: Red Hat Developer Blog
Andrea Veri

MCP servers vs. skills: Choosing the right context for your AI

2026-05-25 07:01
Large language models (LLMs) are powerful when given the right context. The article discusses two methods to enhance LLM capabilities: Model Context Protocol (MCP) servers and skills. MCP servers connect LLMs to external data, streamlining access and ensuring secure data handling. This allows models to retrieve and process real-time information efficiently. On the other hand, skills provide structured instructions for consistent output and specialized tasks. They help LLMs execute repetitive...
Source: Red Hat Developer Blog
Cedric Clyburn

How to route external and local LLMs with Models-as-a-Service

2026-05-25 07:01
AI applications often face challenges when integrating multiple model providers like OpenAI and Anthropic directly into their code. This can complicate switching models and managing requests. To address this, a gateway can be introduced between applications and model providers, allowing for a consistent API while handling routing separately. With Red Hat OpenShift AI 3.4, the Models-as-a-Service (MaaS) capability simplifies this process, offering centralized governance for both self-hosted...
Source: Red Hat Developer Blog
Edward Arthur Quarm Jnr

Protect data offloaded to GPU-accelerated environments with OpenShift sandboxed containers

2026-05-22 07:01
🚀 The rise of AI is reshaping data security in GPU-accelerated environments. Organizations are increasingly concerned about protecting sensitive data and code during computations. Confidential computing technologies, like AMD SEV-SNP and Intel TDX, are essential for creating trusted execution environments (TEEs) that secure memory access. NVIDIA’s confidential GPUs extend these protections to the GPU level, enabling secure workloads in shared infrastructures. Key features include device...
Source: Red Hat Developer Blog
Claudio Carvalho, Pradipta Banerjee, Pei Zhang

Case study: Measuring energy efficiency on the x64 platform

2026-05-22 07:01
In our latest case study, we analyzed a 32-core x64 system with a dual-port 100 GbE network card. Key focus areas included throughput measurement, CPU utilization, computational efficiency, and power consumption. The testing methodology followed our previous blog on optimizing energy efficiency. We observed that while single-core performance is solid, multicore scaling shows diminishing returns as resources are contested. Additionally, hidden hardware limitations were noted, impacting...
Source: Red Hat Developer Blog
Adam Okuliar, Otto Sabart

How to prevent AI inference stack silent failures

2026-05-22 07:01
To ensure reliable AI performance in production, it's crucial to implement an API layer between your application and the inference engine. This setup helps manage state and observability, but silent failures can still occur. Running an end-to-end benchmark, like the Berkeley Function-Calling Leaderboard (BFCL), is essential to identify these issues. Testing with the latest versions of OGX and vLLM on OpenShift AI 3.4 resulted in notable accuracy improvements. For more details on setup and...
Source: Red Hat Developer Blog
Bill Murdock, Robin Narsingh Ranabh

A guide to JIT checkpointing with Kubeflow Trainer on OpenShift AI

2026-05-21 07:16
🚀 Learn about Just-In-Time (JIT) checkpointing with the Kubeflow Trainer on OpenShift AI! This guide details how to implement resilient model checkpointing to protect against interruptions during training. It covers both Persistent Volume Claim (PVC) and S3 storage options. Follow the step-by-step instructions and optimize your training process! 📊🔍 #Kubeflow #OpenShiftAI #MachineLearning #DataScience #Checkpointing
Source: Red Hat Developer Blog
Esa Fazal, Hari Haran Rathinakumar

How to manage TLS certificates used by OpenShift GitOps operator

2026-05-21 07:16
🔒 The latest Red Hat OpenShift GitOps operator 1.20.2 enhances TLS certificate management for Argo CD. It supports both the Mozilla CA bundle for widely trusted sources and allows admins to pin additional certificates through a config map. This flexibility enables a mix of public and internal repositories. Real-world challenges include managing wildcard certificates and indirect TLS connections, which Argo CD doesn't automatically handle. Discover the balance between strict certificate...
Source: Red Hat Developer Blog
Oliver Gondža

Configure a split disk on OpenShift Container Platform

2026-05-21 07:16
🚀 Managing disk space is crucial for AI/ML workloads on Red Hat OpenShift. The new **split disk configuration** directs newly pulled container images to a separate filesystem, keeping the boot disk focused on runtime data. This setup enhances disk space allocation and is particularly useful for environments running large images. The article guides through the configuration process for OpenShift 4.22 on AWS and GCP. For optimal performance, consider this developer preview feature! #OpenShift...
Source: Red Hat Developer Blog
Neeraj Krishna Gopalakrishna