TLDR DevOps 2025-10-20

Managing glue code / internal automations / dashboards? you can do better than fragile Cron jobs or toy automation tools. Ondeva is workflow automation for devs, not hobbyists:

• Built for reliability: Replaces ad-hoc scripts with logged, testable workflows. Every execution is visible, every failure recoverable.
• Reclaim dev hours: Ops and end users can build safe automations with guardrails, so you can get back to real engineering work.
• Centralize automations: Connect any system or database directly. Normalize, clean, and map data without writing ETL scripts.
• Scales with you: Ondeva scales automatically - no infra setup, no server babysitting. 99.9% reliability, enterprise-ready

Book a workflow call >

Pulumi ESC now offers a redesigned onboarding experience and automated OpenID Connect (OIDC) provider setup for AWS, Azure, and Google Cloud. The update simplifies secrets management and configuration orchestration across infrastructure and applications by automating the setup of trust relationships and ESC environments.

Datadog has launched a new OTLP Metrics API, enabling direct ingestion of OpenTelemetry metrics from serverless environments, cloud-provider managed OTel distributions, and applications emitting metrics in OTLP. The API expands observability options by making it easier to capture and analyze telemetry data from environments where collectors can't be deployed, such as Azure Functions, Google Cloud Functions, and Cloudflare Workers. This new API complements Datadog's existing OTel APIs for logs and traces, providing a comprehensive framework for routing OTLP data into the platform.

Cloudflare has introduced Monitor Groups for its Load Balancing service, allowing users to bundle multiple health monitors into a single entity for more accurate application availability assessments. Available via API for Enterprise customers now and coming to the dashboard for all Load Balancing users soon, Monitor Groups enable sophisticated, multi-service health checks by designating critical components and using aggregated health scores for failover decisions. Users can add up to five monitors to a group, designating them as critical or observational and setting a quorum-based threshold to avoid false positives.

Cilium, an eBPF-powered networking solution, helps reduce cross-Availability Zone (AZ) traffic costs on AWS Kubernetes clusters. By using features like topology-aware routing and ENI mode integration, Cilium ensures traffic stays within the same AZ whenever possible. It employs the Linux kernel's eBPF technology to transform Kubernetes networking with efficiency and deep visibility.

GitLab object storage should use a consolidated form with IAM-based authentication, dedicated buckets, server-side encryption, and proxy_download settings optimized for performance. The container registry requires separate configuration with the s3_v2 driver, IAM authentication, encryption, and redirect-based direct downloads to ensure reliability and security.

Running Linux on Azure can cut your infra spend by up to 80%. Microsoft Partner VIAcode is offering zero downtime, zero hassle, zero cost migration of Linux workloads, PostgreSQL, and MySQL to Azure. Read the guide or schedule your free migration assessment

SurfSense, an open-source alternative to NotebookLM and Perplexity, integrates with external sources like search engines, Slack, Jira, and YouTube to create a customizable AI research agent. The tool connects to personal knowledge bases, supports 50+ file extensions, and provides cited answers. It also works with local LLMs via Ollama and is easily deployable.

Stagehand is an AI browser automation framework that allows developers to automate web tasks using both code and natural language. It integrates with Playwright and SOTA computer use models from OpenAI and Anthropic, allowing users to preview and cache actions to save time and tokens.

OpenAI has introduced full Model Context Protocol support in ChatGPT, enabling developers to use custom connectors for both read and write actions through Developer Mode.

Java utilizes garbage collection (GC) to automatically reclaim memory, freeing developers from manual memory management but introducing performance tradeoffs. Modern JVMs use generational GC, dividing the heap into young and old generations, with OpenJDK offering several collectors like Serial GC for small heaps, Parallel GC for high throughput, G1 for balanced performance, and Shenandoah/ZGC for low latency applications. The choice of GC depends on whether the application prioritizes throughput or low latency, with ZGC becoming the default in JDK 23.

The Azure DevOps MCP Server has moved from preview to general availability, featuring improved login, authorization, tooling, and domain support for scoped tool usage. It provides a secure local bridge between AI assistants and Azure DevOps, delivering real-time project context for more accurate responses while maintaining data privacy within local environments.

See how the sidecar model makes your internal tools auth easier to understand, audit, and extend. Read the blog.

Observability is evolving from fragmented tools to unified platforms that drive business outcomes, with organizations progressing through stages of tool consolidation, SLO adoption, and AI-assisted insights.

Argo CD's Pull Request Generator enables ephemeral environments for testing pull requests by watching Git repositories, deploying manifests for matching PRs, and automatically cleaning up resources once the PR is closed.

Amazon EC2 Capacity Manager offers a centralized solution for monitoring, analyzing, and managing capacity usage across all accounts and AWS Regions from a single interface.