Azure Weekly

Issue #564 - 24th May 2026

Highlights this week include: Cloud Native Platforms: Build: Part 1 of a three-part series setting out five engineering disciplines that decide whether a cloud-native platform scales gracefully or accumulates technical debt, starting with flexibility (configuration over conditional logic, feature flags, additive schema versioning) and resilience patterns such as idempotent operations, the transactional outbox, circuit breakers and bulkheads. Cloud Native Platforms: Run: Part 2 argues that platforms are defined by how they are operated rather than how they are built, covering five operational disciplines anchored on observability with user-journey SLIs and burn-rate alerts against SLOs, and treating the incident lifecycle as an engineering contract. Cloud Native Platforms: Evolve: Part 3 looks at how AI is reshaping the whole software lifecycle (planning, design, development, testing, release, operations) rather than just editor autocomplete, with six disciplines for moving from individual assistance to structured, agentic workflows with explicit human checkpoints. Powering multi-cluster workloads with seamless crosscluster networking for Azure Kubernetes Fleet Manager: Introduces a managed Cilium-based cross-cluster networking layer (in preview) for Azure Kubernetes Fleet Manager that lets pods in different AKS clusters talk directly using eBPF routing, with global service discovery via a simple annotation, multi-cluster observability and unified security policies, removing the traditional VPN and gateway "networking tax".

Running a production-like local environment with Aspire: Tim Deschryver shows how .NET Aspire's AppHost lets you model an application's dependencies (databases, queues, blob storage, mail, identity providers) in one place so local development behaves close to production, illustrated with an Azure Blob Storage example that swaps cleanly between Azurite locally and a real Azure Storage account when published. Learn how to host your agents on Microsoft Foundry: A recap of a three-part livestream series on deploying Python AI agents to Microsoft Foundry, covering hosted agents built with Microsoft Agent Framework (with Foundry IQ and multi-agent workflows), agents built with LangChain and LangGraph, and quality and safety evaluations including bulk, scheduled and continuous evals, guardrails and red-teaming, with recordings, slides and open-source code samples.

Finally, Microsoft Fabric variable libraries: best practices guide: Carmel Eve's write-up of a SQLBits 2026 session on Microsoft Fabric variable libraries for managing environment-specific configuration across Dev/Test/Prod, with the key recommendation to set Default values as non-functional placeholders so a missed activation fails loudly rather than silently writing to the wrong environment.