TLDR DevOps 2025-10-03

Build smarter apps with vector search in OpenSearch:

• Low-latency retrieval at scale, open source and production ready
• Ideal for RAG, recommendations, and natural language queries
• Semantic search with dense vector fields and hybrid scoring
• Native support for k-NN and HNSW indexing
• Compatible with your custom ML-generated embeddings
• Integrated with OpenSearch Dashboards for fast tuning and debugging

Skip the black box. Own your search stack with OpenSearch

DigitalOcean has launched Organizations, a new top-level account layer designed to consolidate multiple teams under one cohesive structure, providing centralized control and simplified operations. The new account layer will unify a user's entire DigitalOcean presence into a single group to streamline operations and gain more financial control. With Organizations, users will receive a single invoice, a detailed spending breakdown per team, and centralized user management.

DigitalOcean will implement per-second billing for Droplets starting January 1, aiming to reduce costs for short-lived workloads with a 60-second minimum charge. New dedicated Droplet plans are now generally available for a seamless performance upgrade, while the Bring Your Own IP (BYOIP) feature is generally available to protect IP reputation. Additionally, the VPC NAT gateway is now in public preview, offering centralized egress and static IPs for enhanced security and simplified management.

DigitalOcean announced several enhancements to its storage portfolio, including a Network File System (NFS) generally available on October 20, and Spaces cold storage and usage-based backups, which are now available for public preview. The new NFS service is designed for high-performance AI workloads, while Spaces cold storage offers a low-cost solution for infrequently accessed data, and usage-based backups allow for more frequent backups with flexible retention policies. DigitalOcean's Spaces cold storage will cost users $0.007/GiB per month.

Many organizations still log into each server individually to manage services, which creates bottlenecks and increases risks as infrastructure scales. Ansible's service module enables remote, repeatable, and efficient management of services across Linux and Windows environments, supporting automation of multi-service operations and integration with tools like systemd.

Cloudflare's Automatic SSL/TLS has upgraded over 6 million domains by default to stronger encryption modes, moving sites from Flexible or Off to Full or Full (Strict) without manual intervention. The system now prepares for the post-quantum era by gradually enabling hybrid post-quantum key agreements with origins, improving security and performance while maintaining site reliability.

Enterprises seeking higher performance and flexibility in telemetry data processing are increasingly migrating from Fluentd to Fluent Bit, according to a post highlighting the benefits of the CNCF projects. While Fluentd was initially designed for pushing data to big data platforms using Ruby, Fluent Bit, written in C, offers 10 to 40 times greater log processing performance with the same resources, full OpenTelemetry support, and simpler configuration.

Kosmos, an open-source, all-in-one distributed cloud-native solution, was created to enable the limitless expansion of Kubernetes. Currently, Kosmos primarily consists of three major modules (ClusterLink, ClusterTree, and Scheduler) and is equipped with kosmosctl, a tool for quick deployment of Kosmos components, adding clusters, and testing network connectivity.

Lighthouse provides DNS service discovery to Kubernetes clusters connected by Submariner in multi-cluster environments, working with any CNI plugin.

Meta's Automated Compliance Hardening (ACH) tool combines automated test generation techniques with LLMs to generate mutants for testing and tests that catch those mutants, with privacy engineers accepting 73% of generated tests in trials across Facebook, Instagram, WhatsApp, and wearables platforms from October to December 2024. This technology addresses major barriers to mutation testing deployment by using LLMs to generate fewer, more realistic, and highly specific mutants, while also automatically generating unit tests. Meta is proposing the Catching Just-in-Time Test (JiTTest) Challenge to encourage further innovation in this area.

GKE Autopilot now extends its fully managed, container-optimized compute platform to all qualified clusters, allowing rapid horizontal and vertical scaling, efficient resource use, and simplified cluster operations. Users can run workloads with standard or specialized hardware, set Autopilot as the default for namespaces, and adopt new automatically provisioned compute classes without disrupting existing deployments.

Databricks has created a client-side load balancing system to improve traffic distribution and reduce tail latencies in its Kubernetes clusters, where the default Kubernetes load balancing falls short for high-throughput gRPC connections. The new system uses a custom service discovery control plane to enable intelligent, up-to-date request routing. The system was chosen over alternative approaches like headless services and Istio due to its lightweight, client-driven model.

Don't miss the IBM TechXchange 2025 livestream, featuring exciting product announcements, AI innovations, and enterprise tech insights.

👉 Mark your calendars

Amazon CloudWatch now supports tag-based telemetry, allowing metrics monitoring and alarm management using AWS resource tags to create dynamic views aligned with organizational structures.

Amazon EC2 Auto Scaling now allows forced cancellation of instance refreshes, enabling immediate abortion of in-progress instance launches or terminations.