Hybrid Manager 1.3 release notes v1.3

Release Date: October 10, 2025

We are thrilled to announce the release of EDB Postgres AI Hybrid Manager v1.3.

Release Highlights

• Upgrade reliability: We've successfully hardened platform upgrades across all supported minor and patch releases. This includes a fully validated, self-service pathway for existing EDB Cloud Service (former Big Animal) customers to transition to Hybrid Manager.

• Secured crucial backup and recovery capabilities: Introduced multi-location disaster recovery, simplifying complex switchover and HA management.

• Unified observability: Achieved feature parity for monitoring self-managed bare-metal and CloudNativePG clusters (including Community CNPG), providing a true single pane of glass for all EDB instances.

• Platform stability & flexibility: Validated EDB Cloud Service (formerly Big Animal) to Hybrid Manager migrations, expanded deployment support to SUSE Rancher, and secured crucial backup & recovery flows.

• Data-Sovereign AI for migration: Migration AI Copilot can now use models hosted within Hybrid Manager, ensuring migration queries remain secure and within platform boundaries.

• Expanded AI Factory performance: Introduced VectorChord (vchordrq), a high-performance, GPU-accelerated vector indexing engine for market-leading semantic search and RAG performance.

• Secure AI model access: Enabled secure, authenticated external access to deployed KServe models via a new dynamic proxy layer and centralized AI Model Catalogue management.

New Features and Enhancements

Platform management & reliability

SUSE Rancher platform support: Hybrid Manager is now certified and supported for deployment on SUSE Rancher environments.

Support for EDB Postgres Distributed (PGD) 6.1: Hybrid Manager now supports provisioning and managing clusters running EDB Postgres Distributed (PGD) version 6.1. This upgrade enables customers to utilize the latest features, performance optimizations, and security enhancements provided by the newest PGD release within the unified Hybrid Manager platform.

Decoupled operand lifecycle: We have de-coupled the Postgres operand releases from the main Hybrid Manager release cycle. This foundational improvement enhances platform agility by enabling faster, more frequent, and less disruptive updates to the database engine components.

Local storage support for Engineered System deployments: We have introduced support for provisioning high-performance, low-latency persistent volumes directly from local disks on worker nodes within the Engineered System. This uses the LVM Storage Operator (TopoLVM), enabling robust data persistence for on-premise deployments that do not utilize shared external storage.

Administrator broadcast messaging: Administrators (including project owners) can now broadcast important platform-wide or project-specific messages to all users within Hybrid Manager. This feature is ideal for communicating planned maintenance, outages, or other critical operational updates. These forced notifications are displayed in the user interface via the inbox and information pane to ensure all users are informed.

Disaster recovery and high availability

Multi-location high availability and disaster recovery (DR): Hybrid Manager now supports building highly resilient Postgres architectures across distinct Kubernetes clusters (e.g., across regions) by leveraging CloudNativePG's distributed topology features. This allows users to configure and manage multi-location clusters using native Postgres streaming replication and WAL archiving for robust DR and high availability.

Backup frequency: We've introduced the ability to customize the backup frequency for the Hybrid Manager control plane. Customers can now define the Velero backup schedule directly through the Helm values file, allowing them to precisely control their Recovery Point Objective (RPO) and meet specific operational requirements.

Volume snapshot restore validation: Hybrid Manager now includes enhanced validation to prevent restore operations from stalling and failing due to incompatible storage configurations. The system will now block or issue a clear warning when attempting to restore a cluster from a volume snapshot if the new cluster's storage configuration (e.g., changing from a single data volume to split data/WAL volumes) is not compatible with the source snapshot.

Cross-namespace volume snapshot restore support on RHOS: We have successfully engineered a solution to overcome Red Hat OpenShift's default security context constraints (SCC) that previously prevented restoring a volume snapshot into a different namespace. Hybrid Manager now automatically configures the required security settings to allow cross-namespace restore operations, significantly enhancing disaster recovery and multi-tenancy capabilities for clusters deployed on OpenShift environments. This ensures volume snapshot backups can be reliably restored for data migration or recovery to any target namespace.

Backup support for RHOS: Hybrid Manager now supports volume snapshot-based backup and recovery operations for clusters deployed on Red Hat OpenShift (RHOS) and Engineered System. This provides a faster, platform-native B&R experience for these critical environments.

Upgrade experience

Seamless Hybrid Manager upgrades: We've implemented foundational work to make upgrades between all supported minor and patch releases highly dependable. This enhances platform stability, minimizes the risk of regressions, and provides users with clear expectations and a reliable path for platform transitions.

Transition from EDB Cloud Service: We now support a non-disruptive, self-service path for existing Cloud Service (formerly Big Animal) Bring Your Own Account (BYOA) customers to easily and reliably transition to Hybrid Manager.

Observability, monitoring, and insights

Unified observability: Hybrid Manager supports the collection of essential telemetry and health metrics from self-managed, bare-metal Postgres instances, as well as self-managed CloudNativePG (CNP) and EDB Community CloudNativePG instances, bringing them to parity with HM-managed clusters. The new metrics include crucial operational data such as host metrics, WAL activity, storage utilization, and IOPS statistics. This allows for unified, comprehensive observability within the Hybrid Manager console, regardless of where your Postgres instance is deployed.

Postgres configuration recommendations: Hybrid Manager provides recommendations to help you tune Postgres settings, improve database performance, and optimize resource usage. You can access these suggestions in the Recommendations tab of the cluster’s detailed view.

GenAI workloads with AI Factory

Secure external access to deployed AI models: Hybrid Manager now supports secure, authenticated access to deployed KServe models from outside the EDB Postgres AI Platform. This is achieved by implementing a dynamic proxy and leveraging Istio for robust authentication and authorization (AuthN/AuthZ). This enhancement allows developers to securely integrate deployed AI models (including OpenAI-compatible endpoints) into external applications while maintaining full platform governance.

Centralized AI model catalogue: Hybrid Manager now includes a dedicated model catalogue to serve as the central repository for your organization's AI models. This catalogue provides authorized users with the necessary tools for AI governance and rapid deployment. Users can now list, create, import, edit, and view their entire inventory of AI models. The catalogue integrates directly with KServe, allowing users to deploy an AI model to the model serving infrastructure with a single click from the Hybrid Manager interface.

Dedicated management for AI Model serving clusters (KServe): Hybrid Manager now provides centralized management for the underlying AI model serving infrastructure. Users and administrators can now gain unified visibility and control over deployed KServe clusters and models:

• Model Cluster Overview: A new user interface displays a list of all running model server clusters, along with dedicated detail pages for monitoring and administration.

• REST API Endpoints: Dedicated REST APIs are available for listing running model clusters and retrieving detailed cluster information, facilitating external integration and automation.

VectorChord for high-performance vector indexing: Hybrid Manager now supports and leverages VectorChord (vchordrq), a high-performance vector index engine integrated with Postgres. VectorChord is a scalable, drop-in replacement for the pgvector index, engineered to deliver superior query performance and faster index build times for vector search workloads.

Migrations

Data-Sovereign AI for Migration Portal AI Copilot: AI Copilot can now be configured to exclusively utilize AI Factory Models hosted within Hybrid Manager. This ensures that sensitive migration queries and data remain within the boundary of the platform and are not transmitted to third-party vendors.

Postgres migration assessment details in console: Hybrid Manager's migration assessment capabilities have been expanded to include Postgres as a supported source database. The existing migration assessment details page has been updated to provide Postgres-specific information, offering users a unified and detailed view of your Postgres migration readiness alongside their Oracle assessments.

Agent schema selection and filtering: Users can now precisely control which schemas the migration agent ingests from a source database using the new databases.schema.filter option in the beacon-agent.yaml configuration. This allows platform users to define a subset of schemas to either include or exclude,

Updated Data Migration Service Agent: The underlying components for data migration, previously known as the EDB DMS Reader and Writer, have been unified and updated to the EDB DMS Agent. This change, effective in Hybrid Manager 1.3, provides a streamlined and unified agent for all data ingestion and migration workflows.

Enhanced database migration documentation: We've significantly overhauled all database migration content for improved organization and clarity. All content is now unified under the Migrating databases section, grouped into clear PLANNING, MIGRATING, and REFERENCE categories.

Fixes

• Resolved UI 500 error on database migration page: Fixed an issue where the Migration Databases view would incorrectly return a HTTP 500 error when the number of database items exceeded the default pagination size.

• Resolved cross-version restore issue: Fixed an issue that previously blocked users from selecting the correct Postgres image version during a restore operation following an attempted major version upgrade. Users can now successfully select the original Postgres version image (e.g., v16) and restore a pre-upgrade backup on a new cluster, providing a reliable rollback path if the major upgrade fails or needs to be abandoned.

• Adjusted default log level for beacon-agent sidecar: The default logging level for the beacon-agent sidecar in Postgres pods was unnecessarily set to DEBUG, resulting in excessively noisy logs, especially in GKE deployments. The default log level has been changed to INFO to reduce log clutter and focus on relevant operational information.

• Corrected duplicate metrics after CNP failover: Fixed an issue where connecting external, multi-node CloudNativePG clusters to Hybrid Manager would result in duplicate metric series being displayed on the pg_node level charts following a primary node failover or promotion. Metric collection for these external clusters is now accurately reported after topology changes.

• Eliminated erroneous recommendations for PGD/BDR internal tables: Fixed an issue where the Query Recommendations feature would incorrectly surface suggestions for internal EDB Postgres Distributed (PGD) or BDR management tables. The underlying mechanism (query_advisor extension) has been updated to allow excluding internal system schemas, ensuring that recommendations focus only on end-user application tables and prevent false alarms about internal infrastructure health.

• Initial backup for cluster health: The cluster provisioning workflow has been improved to ensure that a newly deployed cluster is not marked as "Deployed" or "Healthy" in Hybrid Manager until the initial mandatory backup has successfully completed. This change prevents situations you might rely on a seemingly healthy cluster that lacks an immediate restore point.

• Increased minimum resource requirements for EDB Postgres Distributed (PGD): To improve the stability and reliability of PGD deployments, we have increased the minimum required storage and memory allocations. The minimal disk size has been raised to 10 GB (from 1 GB), and the default minimum memory has been set to 2 GB. This change prevents critical operational issues, such as RAFT consensus failures in low-memory environments, and applies consistently across all PGD components, including witness nodes.

• API validation for location-specific networking parameters: Implemented validation in the provisioning API to reject cluster creation requests if networking parameters, such as allowedIpRanges, are specified for a deployment location that doesn't support IP allowlisting. This ensures that users receive immediate feedback on invalid configurations and promotes consistent use of location capabilities.

• Restored tags visibility on the Estate page: Fixed an issue where tags assigned to Postgres clusters were not displayed in the Tags column on the main Estate page. Tags are now correctly retrieved from the API and shown consistently across all relevant user interface views.

• Resolved installation failure on Kubernetes 1.33: Fixed an issue where the transporter component failed to install on Kubernetes version 1.33 due to an incompatibility in the underlying client libraries, ensuring successful installations on modern K8s environments.