AWS

AWS has introduced a user-managed setup (3LO) flow for SharePoint, OneDrive, and Confluence data sources within Amazon Bedrock Managed Knowledge Base. This update removes the previous requirement for admin-level 2LO credential generation, allowing end-users to authenticate directly with their existing third-party accounts. The streamlined process significantly reduces setup time, enabling faster prototyping of AI assistants grounded in corporate documentation.

• End-users can now authenticate directly without needing admin-level access to generate service credentials.
• Setup time is reduced to minutes by leveraging existing third-party accounts via 3LO flow.
• Removes friction for teams prototyping AI assistants on SharePoint, OneDrive, or Confluence data.

AWS has added a native ServiceNow connector to Amazon Bedrock Managed Knowledge Base, enabling direct ingestion of knowledge articles and service catalog items. This feature replaces the previous requirement for custom-built ingestion pipelines by automatically handling crawling, metadata extraction, and incremental synchronization. Engineers can now connect their ServiceNow instances using simple credentials to make institutional knowledge available to AI agents, including access to file attachments.

• Eliminates need for custom ETL pipelines to ingest ServiceNow content into Bedrock.
• Automates incremental syncs, crawling, and metadata extraction for knowledge articles.
• Supports ingestion of service catalog items and associated file attachments.
• Simplifies AI agent access to institutional knowledge via native credential-based connection.

AWS has launched the public preview of Amazon Linux 2027, an OS built on the AL2023 baseline for cloud-native workloads. The release defaults to SELinux enforcing mode and upgrades the kernel to version 7.1 or later. It also integrates AWS-LC for accelerated cryptography and includes drivers for AI accelerators like AWS Neuron.

• Kernel upgrade to 7.1+ requires compatibility checks for existing drivers and modules.
• SELinux enforcing mode is now the default; audit policies before migration to avoid runtime denials.
• AWS-LC integration may impact apps relying on specific OpenSSL ABI behaviors.
• Neuron driver support simplifies deployment for AI/ML inference workloads on AWS hardware.

AWS Gateway Load Balancer can now transmit TCP Reset packets when targets become unhealthy, are deregistered, or idle timeouts expire. This change eliminates the previous fail-open behavior where traffic continued to flow to broken targets, forcing clients to wait minutes for TCP retries. By actively resetting connections, the service allows endpoints to detect failures instantly and reroute to healthy targets within seconds.

• Reduces connection disruption from minutes to seconds during target failures.
• Replaces passive fail-open forwarding with active TCP Reset signaling.
• Applies to unhealthy targets, deregistered instances, and idle timeout expirations.
• No database deep-dive brief present; fleet_impact remains empty.

Amazon Aurora MySQL now supports multi-source replication, allowing a single cluster to ingest data from multiple independent MySQL sources simultaneously. This capability simplifies data consolidation tasks such as shard merging or aggregating regional and departmental databases for centralized reporting. Additionally, delayed replication enables configurable lag on binlog replicas to protect against human error and logical corruption.

• Consolidate data from multiple MySQL sources into one Aurora cluster for unified analytics.
• Use delayed replication as a safety net against accidental data deletion or corruption.
• Simplifies shard merge workflows by aggregating distributed databases centrally.
• Reduces operational overhead for reporting and backup strategies across disparate sources.

Fleet impact: This enables Aurora MySQL to replace complex ETL pipelines for shard merging or regional aggregation, reducing latency and data movement costs. For disaster recovery, delayed replication provides a native, configurable buffer against logical errors, reducing reliance on external point-in-time recovery scripts.

AWS has extended SnapStart support to Lambda functions packaged as container images, significantly reducing cold start latency from several seconds to sub-second levels. This opt-in feature captures a snapshot of the initialized execution environment, allowing large containers with dependencies up to 10 GB to launch instantly. The update helps teams maintain container-based deployment standards without sacrificing the responsiveness required for scalable applications.

• SnapStart is now available for container image functions, not just zip archives.
• Cold starts drop to sub-second, eliminating delays from layer downloads and init.
• Supports larger images up to 10 GB, easing packaging of complex dependencies.
• Opt-in feature requires no infrastructure changes or complex performance tuning.
• Enables container-based deployment standards while maintaining Lambda scalability.