TLDR DevOps 2026-06-08

Blue Pearl's consultant-matching platform had grown in scale and complexity, and staying on Java 11 was starting to hold it back. Deprecated APIs and dependency issues slowed delivery, while missing out on a modern LTS meant weaker security and fewer JVM improvements.

Instead of a typical monthlong uplift, the team moved to Java 25 (LTS) in three days. The work followed a clear sequence: assess the codebase, refactor and align dependencies, update tests, then validate performance and security. The release shipped cleanly, with faster response times and no post-deployment defects.Read how the three-day Java uplift was delivered

Crossplane v1.20.9 introduced a new `crossplane beta upgrade check` command that scans v1.x control planes for breaking changes before upgrading to v2, automatically identifying incompatible resources and providing specific fixes for each issue. The read-only tool addresses a major upgrade hesitation by replacing manual documentation review with automated scanning of compositions, packages, and resources, making it particularly useful since most v1.x control planes can upgrade to v2 without changes, but edge cases exist.

The next generation of AWS Resilience Hub introduces a business-oriented resilience management platform that combines modular resilience policies, application topology modeling, AI-powered failure mode analysis, dependency discovery, and organization-wide reporting. It helps teams define resilience goals, automatically map services and dependencies, identify risks and recovery gaps, and validate compliance at scale across AWS environments through centralized assessments and actionable recommendations.

Amazon Web Services launched a redesigned console experience for Amazon Bedrock that streamlines AI model deployment with support for GPT, Claude, and open-weight models through OpenAI and Anthropic APIs on its new bedrock-mantle inference engine. The new interface features a project-based dashboard with side-by-side model comparison for up to 3 models, integrated API documentation with pre-filled credentials, and AI coding assistant connections, now available across 12 AWS regions, including US East, Europe, and Asia Pacific.

HCP Terraform Projects provide flat workspace organization, shared permissions, variables, and policies for Terraform-centric teams with simpler governance needs. Spacelift Spaces use a hierarchical model with inheritance, granular access control, multi-tenancy, and support for multiple IaC tools, making them better suited for larger organizations with complex infrastructure management requirements.

This guide demonstrates how to load test hosted MCP servers using a reusable Python and Locust harness that faithfully models the MCP lifecycle, including initialization, tool discovery, tool calls, authentication, and session cleanup. The framework supports stateful and stateless MCP servers, multiple authentication patterns, and seamless execution both locally and in Azure Load Testing, enabling teams to measure latency, concurrency behavior, and failure characteristics of production MCP endpoints under realistic AI agent workloads.

Every dependency and automatic dependency update expands a project's supply-chain attack surface, including dev-only tools that still run with access to source code, credentials, and release workflows. Recent compromises like XZ, Trivy, and LiteLLM show why teams should be more skeptical of automatic update pipelines, review dependency changes deliberately, and prefer small amounts of copied code over adding packages for trivial functionality.

Engineers are wiring agents into your clusters faster than anyone can review what they can touch. But “trust the prompt” isn't a security model. Archestra is the open-source control plane for AI that puts deterministic guardrails on every agent action: Kubernetes-native, air-gappable, fully open source. Run it free on GitHub →

Turbovec, a new Rust-based vector search library with Python bindings, can compress a 10-million-document corpus from 31 GB down to 4 GB while searching faster than FAISS by using Google Research's TurboQuant algorithm. The library achieves 16x compression on embeddings and beats FAISS IndexPQ by 0.4–3.4 percentage points on recall metrics across OpenAI's 1536 and 3072-dimensional embeddings, with built-in support for filtered search and stable IDs that survive deletions.

Microsoft open-sourced pg_durable, a PostgreSQL extension that brings durable execution workflows directly into Postgres without requiring external services like cron jobs, queues, or separate orchestration infrastructure. The extension, which is built with pgrx and already integrated into Microsoft's new Azure HorizonDB service, lets developers define multi-step SQL workflows that automatically checkpoint progress and resume from the last successful step after crashes or failures.

According to Docker's State of Agentic AI report, 60% of organizations already have AI agents in production, but 40% cite security and compliance as the top barrier to scaling them further, highlighting the critical need for AI governance frameworks that establish rules, roles, and review processes across the full AI lifecycle. The guide emphasizes that AI governance is no longer optional for organizations using AI at scale, with research from Deloitte showing that companies with strong senior leadership involvement in AI strategy achieve significantly greater business value than those delegating governance solely to technical teams.

Solo.io donated its agentgateway project to the Agentic AI Infrastructure Foundation (AAIF) as a Growth-stage project, positioning it as a unified, Rust-based gateway that handles HTTP, gRPC, MCP, A2A, and LLM traffic while achieving 500k QPS performance in benchmarks. The project, which grew from 100,000 to over 1 million weekly downloads since February and has been adopted by companies including Microsoft, Apple, Adobe, T-Mobile, and Expedia, uses an xDS control plane architecture and draws on lessons from building Istio ambient service mesh to address operational challenges teams face when deploying AI agents that interact with multiple services and APIs.

TOMORROW: Learn how to replace VPN with private, identity-based access across hybrid and multi-cloud - no public endpoints, no rule sprawl. Featuring Twingate's VP Solutions Engineering. Join live / on-demand→

Headlamp offers a migration path that preserves familiar Kubernetes management workflows while adding multi-cluster support, application-centric Projects, plugin extensibility, AI-assisted troubleshooting, and flexible desktop or in-cluster deployment options.

Disconnected feature flag tools and CI/CD pipelines create visibility gaps that slow incident response, complicate audits, increase rollback risk, and reduce release confidence by forcing teams to correlate deployment and feature exposure data across separate systems.