Saturday, August 29, 2026

Pavel Stehule has completed a patch allowing Lua scripts to run within the psql interactive terminal. The implementation enables complex client-side logic, demonstrated by a custom VACUUM wrapper that enhances verbose output readability. A key feature is real-time progress reporting, achieved by polling pg_stat_progress_vacuum every second during long-running maintenance operations.

• psql now supports Lua scripting for advanced client-side automation and UI customization.
• Developers can create custom commands with real-time progress bars for long-running tasks.
• The patch is available via the okbob/lua-psql GitHub repository for testing and integration.

A pg_dump data-only restore fails when unqualified table names in triggers cannot be resolved, even if the tables exist in the schema. This occurs because pg_dump explicitly sets search_path to an empty string at the start of the dump to ensure deterministic behavior. Consequently, any SQL statements referencing tables without a schema prefix, such as those inside AFTER INSERT triggers, throw relation does not exist errors during restoration.

• pg_dump forces search_path to empty string, requiring fully qualified table names.
• Triggers referencing tables by bare name will fail during data-only restores.
• Schema-qualify all table references in triggers to ensure dump compatibility.
• Data-only restores are not immune to schema resolution issues caused by search_path.

Hubert Lubaczewski evaluates two external modules, pg_tre and pg_re2, which bring alternative regular expression engines to PostgreSQL. Using execution plans from explain.depesz.com as test data, he benchmarks these extensions against standard regex capabilities. The analysis highlights potential performance gains and functional differences for complex pattern matching workloads.

• pg_tre and pg_re2 offer alternative regex backends for PostgreSQL beyond the default engine
• Performance varies significantly based on query complexity and data volume
• Benchmarking against real-world execution plans reveals specific optimization opportunities
• These extensions may improve throughput for heavy text processing tasks

Amazon Aurora DSQL now supports foreign key constraints on new and existing tables, allowing developers to enforce relational rules directly in the database. The system rejects writes that would leave a record pointing to a non-existent parent row, ensuring data consistency across the distributed, multi-Region cluster. Administrators can configure deletion and update behaviors using standard options like CASCADE, RESTRICT, and SET NULL.

• Enforce referential integrity natively in Aurora DSQL without application-level checks.
• Supports NO ACTION, RESTRICT, CASCADE, SET NULL, and SET DEFAULT actions.
• Applies to both new and existing tables in compatible AWS Regions.
• Maintains PostgreSQL compatibility while adding strict relational constraints.

Amazon Bedrock in AWS GovCloud (US) now supports SpaceXAI Grok 4.6, a frontier model optimized for coding, agentic workflows, and knowledge work. The model provides a 500k context window and configurable reasoning efforts, accessible via Responses, Chat Completions, and Converse APIs. Customers can leverage cross-Region inference routing across both GovCloud regions for scale, with availability on both bedrock-runtime and bedrock-mantle endpoints.

• Grok 4.6 is now available in AWS GovCloud (US) for secure, compliant AI workloads.
• Supports 500k context window and four levels of configurable reasoning effort.
• Integrates via standard Bedrock APIs: Responses, Chat Completions, and Converse.
• Enables cross-Region inference routing for high-scale deployments in GovCloud.
• Accessible on both bedrock-runtime and bedrock-mantle endpoints.

Amazon SageMaker JumpStart now hosts Meta’s Muse-Glimmer-30B and Alibaba’s Qwen 3.8-27B foundation models. Muse-Glimmer-30B is a dense 30B-parameter model featuring a 1.8B ViT-G/14 encoder for multimodal inputs and a 131K+ context window, optimized for autonomous agentic workflows with tool use and failure recovery. Qwen 3.8-27B complements this with capabilities in multimodal long-horizon reasoning, giving engineers more specialized options for scalable AI solutions on AWS.

• Muse-Glimmer-30B supports autonomous agentic tasks with multi-step reasoning and tool use.
• Model includes a ~1.8B ViT-G/14 perception encoder for interleaved text and image inputs.
• 131K+ context window enables handling of long-horizon reasoning and complex workflows.
• Qwen 3.8-27B adds multimodal capabilities for extended reasoning tasks.
• Both models are now deployable via SageMaker JumpStart for enterprise AI solutions.

NVIDIA has released the Cosmos3-Edge, Cosmos3-Nano, and Cosmos3-Super foundation models on Amazon SageMaker JumpStart. These open, frontier omnimodal world models enable developers to build robots and autonomous systems that perceive, reason, and act in the physical world. Cosmos3-Edge specifically targets on-device control with a 4B-parameter model optimized for real-time visual reasoning at 15 Hz.

• SageMaker JumpStart now hosts NVIDIA's full Cosmos3 family for physical AI development.
• Cosmos3-Edge is a 4B-parameter model designed for real-time robot control and edge inference.
• Models support vision AI tasks requiring perception, reasoning, planning, and action.
• Cosmos3-Edge operates at 640x360 resolution and generates 32 actions per inference at 15 Hz.

Amazon Redshift now connects with the Agent Toolkit for AWS, allowing AI coding assistants like Claude Code and Cursor to manage data warehouses directly. This integration leverages an AWS MCP server to execute authenticated API calls and utilizes curated Redshift skills to guide agents. These skills provide SQL references, metadata discovery, and best practices for data loading and materialized views to reduce errors.

• AI agents can now execute authenticated AWS API calls via the Model Context Protocol server.
• Redshift skills offer curated SQL syntax references to minimize query generation errors.
• Metadata discovery features allow schema exploration without manual SQL writing.
• Guidance covers data loading patterns, materialized views, and extensions like Pivot.

Amazon Bedrock AgentCore Memory now supports fine-grained access control, allowing per-user and per-tenant memory isolation via AgentCore Gateway. This feature enables OAuth (JWT) authentication and Cedar policy enforcement to restrict access based on caller identity and token claims. By moving authorization to the infrastructure layer, developers can enforce namespace-derived restrictions and operation-level permissions without custom application logic.

• Enforce per-user and per-tenant memory isolation using AgentCore Gateway.
• Integrate OAuth (JWT) authentication with Cedar policies for identity-based access.
• Restrict memory records to namespaces derived from user token claims.
• Move access control from app code to infrastructure for better security posture.
• Allow or deny specific Memory operations based on the authenticated caller.

Amazon Bedrock AgentCore Memory now supports flexible namespace variables, allowing developers to scope long-term memories along application-specific dimensions like tenant or environment. This feature eliminates the need for duplicate strategies or overloading built-in variables by enabling fine-grained isolation in multi-tenant setups. Users define up to five keys per memory resource and supply values at runtime via the CreateEvent API, which the service substitutes into namespace templates during memory extraction.

• Define up to five flexible namespace keys per memory resource for custom scoping.
• Scope memories by tenant, org, team, or environment without duplicate strategies.
• Supply namespace values at runtime through the CreateEvent API.
• Variables are substituted into strategy namespace templates during extraction.
• Enables fine-grained access control for complex, multi-tenant applications.

Amazon Bedrock AgentCore is now available in US West (N. California) and Asia Pacific (Hyderabad) regions. This expansion enables engineers to build, connect, and optimize AI agents closer to end users for lower latency. All core capabilities including runtime, identity management, tool connectivity, and observability are supported at launch in these new regions.

• Bedrock AgentCore is now live in US West (N. California) and Asia Pacific (Hyderabad)
• New regions allow building agents closer to users to reduce latency
• Full feature set including runtime, IAM, tool connectivity, and observability is available
• Infrastructure-layer security enforcement remains consistent across the expansion

The Amazon CloudWatch agent now supports direct collection of systemd journal entries from Linux instances, eliminating the need to buffer logs to disk files first. This update aligns with modern distributions like Amazon Linux 2023, which default to journald instead of traditional text logs. The agent preserves structured metadata, such as the systemd unit, during ingestion to CloudWatch Logs.

• Eliminates disk I/O overhead from exporting journald to temporary files before ingestion.
• Preserves rich structured metadata like systemd units for better log context.
• Simplifies configuration on Amazon Linux 2023 and other journald-first distros.
• Reduces agent complexity by removing manual export step requirements.

Uber has introduced GitFarm, a centralized service that executes Git operations for large-scale monorepos, removing the need for local repository clones. The platform leverages prewarmed checkouts, ephemeral sandboxes, and repository synchronization to optimize resource usage. Additionally, it employs gRPC streaming to significantly reduce startup latency for automation services managing thousands of repositories.

• Eliminates local clones by centralizing Git operations as a service for monorepos.
• Uses prewarmed checkouts and ephemeral sandboxes to lower resource consumption.
• Reduces startup latency for automation pipelines via gRPC streaming.
• Scales efficiently across thousands of repositories without local storage overhead.

The original Twelve-Factor manifesto has been updated for 2025 to reflect current cloud-native practices. The revision addresses changes in deployment models, infrastructure abstraction, and developer workflows since the initial release. It serves as a refreshed guide for building software-as-a-service applications in today's distributed environments.

• Core principles remain relevant but are contextualized for modern cloud platforms.
• Updates reflect shifts in deployment, infrastructure, and developer tooling.
• Serves as a baseline for building SaaS apps in 2025 and beyond.
• No database-specific deep dive provided; applicable to general app architecture.

Percona Server for MongoDB now includes a Software Bill of Materials to enhance supply chain transparency. This machine-readable document enumerates all incorporated components, including transitive dependencies, to aid in license compliance and vulnerability detection.

• SBOMs provide machine-readable visibility into all direct and transitive dependencies.
• Facilitates automated license compliance checks across the software stack.
• Enables faster detection and mitigation of known vulnerabilities in upstream libraries.
• Improves overall security posture by clarifying the third-party component inventory.

Amazon EC2 C8gn instances powered by Graviton4 processors are now available in the Europe (Paris) region. These instances deliver up to 30% better compute performance than the previous C7gn generation and utilize 6th generation Nitro Cards. They support up to 600 Gbps network bandwidth, making them suitable for network-intensive workloads like AI inference and data analytics.

• Graviton4 chips in C8gn offer 30% compute uplift over C7gn for better efficiency.
• 6th gen Nitro Cards enable 600 Gbps bandwidth, the highest among network-optimized instances.
• Ideal for network-heavy tasks including AI/ML inference and network virtual appliances.
• Expansion to Paris region allows EU-based teams to leverage these new capabilities locally.

AWS Glue 5.1 is now available in the AWS European Sovereign Cloud Region, upgrading core engines to Apache Spark 3.5.6, Python 3.11, and Scala 2.12.18. The release enhances open table format support, including Apache Hudi 1.0.2, Iceberg 1.10.0, and Delta Lake 3.3.2. Notably, it introduces Iceberg format version 3.0, adding features like default column values, deletion vectors, and row lineage tracking.

• Glue 5.1 is now live in the AWS European Sovereign Cloud Region for compliant workloads.
• Core engines updated to Spark 3.5.6, Python 3.11, and Scala 2.12.18.
• Open table formats updated: Hudi 1.0.2, Iceberg 1.10.0, Delta Lake 3.3.2.
• Iceberg 3.0 support adds deletion vectors, default column values, and row lineage.
• Lake Formation fine-grained access control extended to write operations.

AWS has expanded Capacity Reservation Resource Groups to include Amazon EC2 Capacity Blocks for ML and interruptible Capacity Reservations, in addition to the previously supported On-Demand Capacity Reservations. This enhancement allows practitioners to target multiple reservation types within a single group when launching instances via EC2 Fleet or Auto Scaling. The change simplifies capacity management across diverse reservation portfolios by enabling unified targeting strategies.

• CR Resource Groups now accept ML Capacity Blocks and interruptible reservations, not just ODCRs.
• Unified targeting allows launching instances across mixed reservation types in a single request.
• Simplifies capacity planning for workloads requiring specific ML hardware or flexible pricing models.
• Compatible with EC2 Fleet and Auto Scaling groups for automated capacity provisioning.

A PostgreSQL standby experienced unexpected replication lag that escalated to AWS Support for investigation. The root cause was identified as the hidden single-flow bandwidth limit inherent to the EC2 instance type. This bottleneck constrained throughput on the single network connection used for storage traffic, despite adequate aggregate bandwidth availability.

• Single-flow bandwidth limits on EC2 can throttle FSx traffic, causing replication lag.
• Aggregate bandwidth metrics may mask underlying single-connection bottlenecks.
• Verify EC2 instance network performance specs when troubleshooting storage latency.
• Consider multi-path or enhanced networking configurations to bypass single-flow limits.

Fleet impact: For DBAs managing PostgreSQL on AWS FSx, check your EC2 instance's single-flow TCP bandwidth limits if replication lag spikes without storage saturation. If the instance type has low single-flow caps, consider upgrading to an instance with higher per-flow limits or using multiple mount targets to distribute traffic across separate flows.

AWS Backup now allows cross-Region backup copying and logically air-gapped vault storage for Amazon DocumentDB in nine additional regions across Asia Pacific, Europe, Africa, and Israel. Users can configure backup plans or on-demand jobs to move DocumentDB backups into and out of these new locations. The air-gapped vaults provide immutable, isolated storage with default locking and encryption options to support strict compliance and disaster recovery needs.

• Cross-Region copy for DocumentDB is now available in nine new AWS regions.
• Logically air-gapped vaults offer immutable, isolated backup storage in these regions.
• Supports disaster recovery, business continuity, and compliance requirements.
• Backups can be shared for recovery using AWS Resource Access Manager (RAM).

Mountpoint for Amazon S3 now supports explicit memory limits, allowing operators to cap usage either via a defined target or automatic environment detection. This feature addresses previous issues where memory consumption could expand over time, potentially destabilizing co-located memory-intensive workloads like ML training. By reserving memory for companion applications, the update ensures stable performance in shared environments such as Amazon EKS.

• Define explicit memory targets for Mountpoint to prevent resource contention.
• Enable automatic memory detection for safe defaults in dynamic environments like EKS.
• Mitigate stability risks when running S3 mounts alongside heavy analytics or ML workloads.
• Prevent unbounded memory growth that previously degraded performance over time.

Amazon RDS for Oracle now offers the July 2026 Release Update for versions 19c, 21c, and 26ai, including critical security patches. For Oracle 19c, the RU naming convention has shifted to a new format that explicitly denotes the release and maintenance patch release dates. Oracle 21c continues to use its existing naming structure for RUs.

• July 2026 RU is available for Oracle 19c, 21c, and 26ai on RDS.
• 19c RU naming changed to <version>.ru-<YYYY-MM>.mrp-<YYYY-MM>.r<N> format.
• 21c RUs retain their original naming convention without changes.
• Monthly CSPUs for 19c and 26ai will be bundled as MRP versions.
• Upgrade recommended to apply included security updates immediately.

Fleet impact: Update Oracle 19c instances to the new July 2026 RU to align with the new naming scheme and apply security fixes. Verify that automation scripts parsing the version string are updated to handle the new <version>.ru-<YYYY-MM>.mrp-<YYYY-MM>.r<N> format to prevent deployment failures.

Google has released version 1.18.0 of GoogleTest, the combined testing and mocking framework for C++. This update enforces a minimum C++17 standard requirement. The project is also planning to introduce a dependency on Abseil in future releases.

• Upgrade to C++17 or later to build this version of the framework.
• Expect an incoming Abseil dependency affecting build configurations.
• Use the GitHub Pages-hosted docs instead of local repo files.
• Maintains unified GoogleTest and GoogleMock release cycles.

JetBrains released a curated set of coding guidelines specifically designed to instruct AI agents on writing idiomatic Go. The rules prioritize built-in functions and recent language features, such as max, slices.Contains, and errors.AsType, over verbose manual implementations. By analyzing the go.mod file, agents can dynamically apply the appropriate standards for the project's specific Go version, ensuring consistency with the modernize analyzer.

• AI agents use these rules to replace manual loops with slices.Contains and if-else chains with max/cmp.Or.
• Guidelines enforce type-safe error handling via errors.AsType and pointer creation with new() where applicable.
• Agents detect the target Go version from go.mod to limit feature usage to supported releases.
• Covers features from Go 1.0 through 1.27, aligning with the modernize analyzer's scope.
• Reduces boilerplate in AI-generated code by enforcing standard library idioms over custom logic.