Sunday, September 27, 2026

TypeSafe’s migration to store AI-generated labels in PostgreSQL revealed that interleaved batch updates caused significant bloat. Range-based batching at default fillfactor nearly doubled heap size and index weight, while higher fillfactor with frequent VACUUM reduced heap growth but still increased index size. Only 11.5% of updates were HOT, indicating heavy index maintenance overhead during the backfill process.

• Interleaved batch updates cause disproportionate WAL generation and heap bloat in PostgreSQL.
• Range-based backfills at default fillfactor significantly increase both heap and index sizes.
• Increasing fillfactor and running VACUUM between batches reduces heap bloat but not index growth.
• Low HOT update rates indicate heavy index maintenance costs during large-scale backfills.
• Monitor index size growth when planning large table migrations or label storage updates.

PgBouncer 1.26.0 addresses three critical denial-of-service vulnerabilities, including crashes from malformed SCRAM messages and infinite loops from integer overflows. The release introduces new configuration options like pool_idle_timeout and per-user query_wait_timeout, while also tracking search_path and transaction read-only settings by default. Additionally, it adopts Meson for building and removes the deprecated online restart functionality.

• Patch immediately to prevent DoS from unauthenticated clients exploiting SCRAM or buffer logic flaws.
• Update build pipelines to support Meson, as legacy build methods are being phased out.
• Leverage new pool_idle_timeout and granular query_wait_timeout for better resource control.
• Verify that search_path and default_transaction_read_only tracking aligns with current app expectations.

Amazon RDS for PostgreSQL 19 Beta 4 is now available in the Database Preview Environment for evaluation. This release refines query performance and autovacuum management introduced in earlier betas, alongside stability improvements. Key fixes address pg_stat_autovacuum_scores reporting for TOAST tables and correct freeze-score scaling for better tuning.

• Preview environment allows early testing of PostgreSQL 19 Beta 4 features on RDS.
• Autovacuum parallel workers now share a rebalanced cost limit to respect configured limits.
• Fixed pg_stat_autovacuum_scores view to properly report on TOAST tables.
• Corrected freeze-score scaling enables more effective autovacuum prioritization tuning.
• Refer to community announcement for full technical details on this beta release.

AWS has launched a Kubernetes-native inference gateway for SageMaker HyperPod that replaces round-robin load balancing with real-time signal-driven routing. Deployed as an EKS add-on, it uses an Envoy endpoint to route requests based on model names within the payload, eliminating the need for client-side changes. Early metrics indicate significant latency improvements, with p99 Time-to-First-Token reductions of up to 98% in mixed-hardware scenarios.

• Zero app changes required; deploys as an EKS managed add-on on existing HyperPod infrastructure.
• Routes by reading model names from request bodies, allowing single endpoint for multiple models.
• Achieves up to 82% reduction in first-token latency and 98% p99 TTFT drop in burst traffic.
• Built on Envoy for HTTPS termination and GPU-aware routing to optimize mixed-hardware pools.

AWS has made OpenAI's GPT-6 Sol and GPT-6 Luna models generally available on Amazon Bedrock, expanding the GPT-6 family alongside Astra. Sol targets complex recurring tasks and development with significantly improved factuality, while Luna offers high efficiency for volume-heavy operations like summarization and routing. Both models support a 1 million token context window and are optimized for production-scale inference.

• GPT-6 Sol improves factuality by roughly 50% compared to GPT-5.6 Sol for complex dev tasks.
• GPT-6 Luna is optimized for high-volume, focused workloads like extraction and classification.
• Both models support 1M token context windows for long-document processing.
• Bedrock provides the underlying security, scale, and inference engine for production use.

AWS GovCloud (US) now supports Claude Opus 5.5, the latest iteration of Anthropic's top-tier model. This release emphasizes improved efficiency, requiring fewer tokens and offering lower pricing compared to its predecessor. Designed for complex coding and knowledge work, it features adaptive reasoning and detailed progress reporting.

• Claude Opus 5.5 is now live in AWS GovCloud (US) for compliant workloads.
• The model reduces token usage and cost per token compared to Opus 5.
• Cheaper cache reads further lower inference expenses.
• Optimized for long-running coding and knowledge synthesis tasks.
• Provides structured feedback on actions taken and next steps.

AWS has introduced native data source connectors for Salesforce and Zendesk within Amazon Bedrock Managed Knowledge Base. These integrations automate data crawling, metadata extraction, and incremental synchronization, eliminating the need for custom ingestion pipelines. By simply providing instance credentials, engineers can ground AI agents in existing support and product knowledge.

• Native connectors replace custom ingestion pipelines for Salesforce and Zendesk data.
• Automated crawling and incremental syncs keep the knowledge base current.
• Metadata extraction is handled automatically by the service.
• Enables rapid development of support bots grounded in existing ticket and article data.

NVIDIA has released Model-Optimizer, a unified library designed to compress and accelerate deep learning models. It supports inputs from Hugging Face, PyTorch, and ONNX, applying techniques like quantization, distillation, and neural architecture search. The optimized models are intended for deployment in downstream inference frameworks such as TensorRT-LLM, TensorRT, and vLLM.

• Supports Hugging Face, PyTorch, and ONNX model inputs for broad compatibility.
• Integrates quantization, pruning, distillation, and speculative decoding in one tool.
• Targets faster inference via TensorRT-LLM, TensorRT, and vLLM deployment.
• Reduces model size and compute requirements for production AI workloads.

AWS has extended Lambda durable functions to the European Sovereign Cloud region, enabling reliable multi-step application orchestration. This update allows developers to use primitives like step and wait to checkpoint progress and pause executions without incurring compute charges. The availability supports complex workflows such as AI tasks and order processing while maintaining local compliance and security requirements.

• Lambda durable functions are now available in the European Sovereign Cloud region.
• New primitives enable checkpointing and recovery without incurring compute charges during pauses.
• Supports compliance-sensitive orchestration for AI workflows and order processing.
• Activates durable function capabilities for Python 3.13 and other supported runtimes.

Amazon Connect Customer now enables AI agents to collaborate with each other and external agents using the open A2A protocol over text or bidirectional voice. This allows frontline agents to delegate complex tasks to specialists, such as fraud assessment agents, and transfer disputes directly to resolution agents. The capability expands the Linux Foundation's A2A protocol to include voice interactions for real-time cooperation.

• Agents can join live interactions to handle specialized tasks like fraud checks or dispute resolution.
• Supports both text and bidirectional voice collaboration via the open A2A protocol.
• Enables delegation from frontline agents to specialist AI models in real time.
• Extends Linux Foundation A2C protocol to include voice for richer agent cooperation.

Amazon CloudWatch Omni is now generally available as an AI-powered evolution of CloudWatch, designed to centralize observability for teams, applications, and agents. It merges OpenTelemetry interoperability with CloudWatch’s scale, offering a unified view across AWS, Azure, and multi-account environments. The platform provides both a standalone web interface with SSO and a local IDE extension for direct agent troubleshooting.

• Centralizes telemetry across AWS, Azure, and multi-account setups into unified 'spaces'.
• Supports AI agents natively with automatic service discovery and dependency mapping.
• Offers a local IDE extension for hands-on agent debugging alongside the web dashboard.
• Combines OpenTelemetry interoperability with CloudWatch’s underlying reliability and scale.

A user attempted to run pgBackRest on a standby after promotion with archive_mode disabled, expecting the tool to handle the state. Even with archive-mode-check=n, the backup failed because the underlying PostgreSQL server enforces archive_mode for WAL archiving operations. Modifying pgBackRest source code bypassed the error, but the resulting backup lacks the necessary WAL segments for point-in-time recovery. The author concludes that relying on unsupported workarounds is risky and recommends restoring a valid configuration via restart or switchover.

• pgBackRest cannot reliably back up a standby that lacks active WAL archiving.
• archive-mode-check=n only bypasses the tool's validation, not PostgreSQL's enforcement.
• Bypassing checks creates backups that may be incomplete or unrecoverable.
• Always ensure archive_mode=on before attempting pgBackRest operations on standbys.
• Use controlled switchover or restart to restore supported backup configurations.

Amazon EventBridge has relaunched its Custom event bus to support enterprise-scale event-driven architectures. The updated serverless broker now supports strict message ordering, built-in event retention, and open event formats to better decouple teams. Publishers can share these buses across multiple accounts via AWS Resource Access Manager for centralized routing.

• New Custom event bus supports strict ordering for critical event sequences
• Built-in retention eliminates need for external storage for event history
• Share buses cross-account via Resource Access Manager for centralization
• Supports advanced event transformations and open event formats

OpenBao is an open-source software solution designed to manage, store, and distribute sensitive data such as secrets, certificates, and keys. It serves as a community-driven alternative to HashiCorp Vault, operating under the OpenSSF. The project emphasizes security and user trust, providing dedicated channels for responsible disclosure of vulnerabilities.

• OpenBao provides a transparent, open-source path for secrets management without licensing restrictions.
• Supports core use cases: secret storage, certificate issuance, and key distribution.
• Security issues must be reported responsibly to the OpenSSF mailing list.
• Active working groups focus on scalability, PKCS#11, UI, and supply chain.
• Direct replacement for Vault for teams requiring open-source compliance.

AWS IAM outbound identity federation now allows access to OIDC discovery and JWKS endpoints through interface VPC endpoints using AWS PrivateLink. This change enables workloads to retrieve short-lived JWTs and verify them without routing traffic over the public internet. It enhances security by keeping identity federation metadata within the VPC boundary.

• Route OIDC discovery and JWKS traffic through AWS PrivateLink to avoid public internet exposure.
• Maintain short-lived JWT workflows for external service access without long-lived credentials.
• Reduce attack surface by keeping identity federation metadata within the VPC boundary.
• No public internet gateway or NAT required for OIDC metadata retrieval from within VPC.

Amazon RDS Custom for SQL Server now supports the latest Cumulative Update, specifically SQL Server 2022 CU26 (KB5093420). This update maps to RDS version 16.00.4265.3.v1 and includes various improvements and fixes detailed in Microsoft's documentation. Administrators can apply these patches via the AWS Management Console, CLI, or SDKs.

• RDS Custom for SQL Server now supports CU26 KB5093420 (RDS v16.00.4265.3.v1).
• Patches include Microsoft-recommended fixes and improvements for SQL Server 2022.
• Upgrade instances using the RDS Console, AWS CLI, or AWS SDKs.

Amazon EC2 M8i and M8i-flex instances are now live in the AWS European Sovereign Cloud (Germany) region. Powered by custom Intel Xeon 6 processors exclusive to AWS, these instances deliver significantly higher memory bandwidth and performance compared to previous generations. The new hardware offers up to 20% better overall performance than M7i instances, with specific gains for database and AI workloads.

• M8i/M8i-flex now available in AWS European Sovereign Cloud (Germany)
• Custom Intel Xeon 6 processors provide 2.5x memory bandwidth vs prior gen
• Up to 30% faster for PostgreSQL and 40% faster for AI deep learning models
• 20% general performance uplift over M7i with 15% better price-performance

Amazon EMR on EKS now supports interactive Apache Spark sessions via Spark Connect, allowing developers to run persistent Spark contexts from local IDEs or SageMaker Unified Studio. This architecture decouples the application client from the Spark driver, enabling seamless blending of local Python execution with remote Spark operations on existing EKS clusters. The feature is designed to improve developer productivity by maintaining preferred tooling while leveraging managed compute resources.

• Developers can debug Spark apps interactively from local IDEs like VS Code or Jupyter.
• Spark Connect architecture decouples client code from the remote Spark driver.
• Persistent Spark contexts span across cells and scripts for iterative development.
• Integrates with Amazon SageMaker Unified Studio for managed notebook workflows.

Amazon EMR now offers 36-month support for designated LTS releases, starting with emr-spark-8.1.0 and Apache Spark 4.1. This release includes critical security and bug fixes to help teams stabilize production workloads without immediate upgrade pressure. It also introduces full support for Apache Iceberg v3, enabling advanced features like geospatial queries and schema evolution.

• LTS releases get 36 months of support for critical and high severity issues.
• Apache Iceberg v3 support brings geospatial and high-precision timestamp capabilities.
• Spark SQL can now reference catalogs by name, including cross-account S3 Tables.
• Automatic detection works for Apache Iceberg, Delta Lake, and Apache Hudi.

Amazon DynamoDB Global Tables with Multi-Region Strong Consistency now supports five new regions, including Canada Central, Europe Stockholm, Europe Spain, Asia Pacific Mumbai, and Asia Pacific Singapore. This brings the total supported regions to 15, allowing teams to configure global tables across three regions in any combination of replicas and witnesses. The expansion enables cross-continent setups that maintain zero recovery point objective while placing endpoints closer to users in more geographies.

• MRSC now covers 15 regions total, adding Canada, Europe, and Asia Pacific locations.
• Supports 3-region configurations with mixed replica and witness roles across continents.
• Enables zero RPO strong consistency for global applications with local read latency.
• Allows routing traffic to nearest endpoints in newly supported regions like Mumbai and Singapore.

pgsql-test addresses the gap in testing database logic by enabling rapid, automated feedback loops similar to those in application code. It allows developers to verify triggers, foreign keys, and row-level security policies without relying on mocks or manual checks. This tool helps keep enforcement logic within Postgres rather than shifting it to the application layer.

• Mocks fail to validate actual DB behavior like triggers or RLS policies.
• pgsql-test enables fast iteration for database-specific logic.
• Reduces architectural pressure to move constraints to app code.
• Supports verifying queries run under correct database roles.