EDB Postgres AI Hybrid Manager#
EDB Postgres AI Hybrid Manager (HM) is the control plane for Postgres, sovereign AI, and analytics workloads across Kubernetes-based infrastructures. It provides a consistent and unified way to provision, operate, and monitor PostgreSQL clusters—whether you’re deploying on-prem, in the cloud, or across hybrid environments.
Hybrid Manager enables end-to-end data workflows from database deployment to AI analytics, all governed within your infrastructure.
Why Hybrid Manager?#
Modern data platforms require more than just database provisioning—they need to support:
Kubernetes-native scalability and orchestration
Resilient, high-availability deployments
AI and GenAI integration
Postgres-native analytics with open formats
Governance, observability, and access control
Hybrid Manager was created to meet these demands. It extends Kubernetes capabilities with built-in Postgres expertise, making it easier to operate stateful workloads without sacrificing the flexibility of containerized infrastructure.
Key capabilities#
Area |
Description |
|---|---|
Database provisioning |
Deploy highly available PostgreSQL or Distributed Postgres clusters using the console, templates, or declarative manifests. |
AI integration |
Run Troubleshooting , serve models with KServe, and build AI knowledge bases governed by your infrastructure. |
Analytics |
Offload cold data to Delta Lake or Iceberg , query with Postgres syntax, and enable hybrid transactional/analytical processing. |
Monitoring and observability |
Use the integrated dashboards that leverage Prometheus, Grafana, and Loki to obtain deep visibility into query health, resource usage, and system alerts. |
Security and access management |
Integrate with identity providers, define RBAC for projects and clusters, and manage secure access to your databases and services. |
Database migration |
Migrate schemas and data from external databases (Oracle, self-managed Postgres) to HM-managed clusters. Leverage integrated services for migration and schema assessment, data transfer (snapshot), and continuous replication (streaming) to achieve minimal downtime migrations. |
Supported environments#
Hybrid Manager runs on Kubernetes and supports these common managed Kubernetes cloud services:
Microsoft Azure Kubernetes Service (AKS)
Amazon Elastic Kubernetes Service (EKS)
Red Hat OpenShift Service on AWS (ROSA)
Google Kubernetes Engine (GKE)
HM also supports on-premises Kubernetes platforms:
Red Hat OpenShift (RHOS)
SUSE Rancher RKE2
It supports both cloud-native and on-premises architectures, allowing you to build sovereign, compliant, and flexible data infrastructures.
Integrated solutions#
Hybrid Manager unlocks core functionality across the EDB Postgres AI ecosystem. It consists of:
Getting Started with AI Factory – Build and run GenAI workloads using private data.
Analytics in Hybrid Manager – Enable open-format analytics with Iceberg, Delta Lake, and tiered offloading.
Simple Migration Agent – Accelerate adoption of Postgres with in-platform migration tools.
Monitoring execution – Monitor operations using built-in observability stack with Kubernetes and database insights.
Missing notifications for task and cluster lifecycle events – Create templates, perform upgrades, configure HA settings, and restore data.
Learn more#
Get started#
Use Hybrid Manager
- Hybrid Manager architecture
- Supported database types
- Known issues
- Multi-DC
- Core platform and resources
- Cross-stream upgrade issues for OpenShift users
upm-beacon-agentmemory limits are insufficient in complex environments- IdP configuration fails after upgrade to version 2026.2
accm-serverfails to start due to missing Langflow secret- No
kapp-controllersupport in Hybrid Manager app to pull images - Beacon server enters a crashloop when spire is broken
- SPIRE agent fails to start on RKE2 with SELinux enabled
- Image library discovery fails when duplicate operand images share the same tags and SHA
- Database cluster engine
- Failure to create 3-node PGD cluster when
max_connectionsis non-default - AHA witness node resources are over-provisioned
- HA clusters use
verify-cainstead ofverify-fullfor streaming replication certificate authentication - Second node is too slow to join large HA clusters
- EDB Postgres Distributed (PGD) cluster with 2 data groups and 1 witness group not healthy
- AKS deployments limited to one PGD data group with private access per region
- Network access type defaults to
Privatewhen not explicitly selected on RKE - CNPG operator enters an infinite reconcile loop when bootstrap
Jobcreation is interrupted
- Failure to create 3-node PGD cluster when
- Backup and recovery
- Replica cluster creation fails when using volume snapshot recovery across regions
- Volume snapshot restoration is limited to the same region
- WAL archiving is slow due to default parallel configuration
transporter-dbdisaster recovery (DR) process may fail due to WAL gaps- Full cluster recovery fails with
pg_ctl: server does not shut downerror
- Agent Factory and model management
- Failure to deploy
nim-nvidia-nvclipmodel with profile cache - AI Model cluster deployment stalls if object storage path for model profiles is empty
- Model configuration settings reset after Innovation Release upgrade
- Error page appears when editing knowledge base credentials
- Missing aidb_users role on PGD witness node prevents AI functionality
- GenAI Builder structures and tools execution failure
- Pipeline Designer-created knowledge bases not visible to GenAI Builder
- Editing external inference service parameters returns a name change error
- Missing validation message for invalid external model names
- Pipeline creation fails on PGD clusters with AIDB v6.1
- External cloud-hosted OCR models fail to register via the HM console
- Failure to deploy
- Analytics and tiered tables
- HM console and observability
- False positive backup alerts for Advanced High Availability and Distributed High Availability (PGD) clusters
- Event markers overlapping when displaying two different events
- HTTP 431 “Request Header Fields Too Large” error when accessing the Estate page
- Missing notifications for task and cluster lifecycle events
- WAL storage percent metric displays incorrect values on the monitoring page
- GUC settings aren’t applied until after a restart
- Alert threshold updates intermittently fail to take effect
- Asset Library apps require Project Owner or Project Editor role
- Asset Library apps sorted by package name instead of display name
- Metabase fails to deploy when using an EPAS cluster as its database backend
- Asset Library apps pgAdmin4 and Apache Superset fail to deploy on RHOS
- Editing cryptographic keys for Airflow or Superset causes the app to enter a failed state
- Superset deployment fails with
DuplicateTableerror - Asset Library launch app always uses port 443 regardless of custom portal port
- Asset Library app CPU and memory metrics include failed and completed pod states
- Deleting an app or flow gets stuck due to missing service account
- Asset Library apps may get stuck pending on RHOS when storage classes differ
- Clusters and upgrades
- Agent Factory
- Knowledge base data isn’t replicated to non-primary nodes on PGD clusters
- Index analysis actions fail due to missing
hypopgextension - Deleted clusters still appear in AI components
- Missing validation for required MCP tool parameters
- Some EDB Component tools are non-functional due to incompatible API request formatting
- Advanced high Availability and Distributed High Availability (PGD) witness groups appear in database selection lists
- Text search may match on non-visible fields
- Langflow MCP server components not supported in HM Flow Hosting
- EDB Knowledge Base component returns a 403 error when dragged into Langflow
- EDB Knowledge Base component fails to connect when using a global variable for the database user
- Running a test query against an empty knowledge base returns a 500 error
- Flow Hosting deployment fails intermittently with
FlowValidationError - EDB Model Server component causes agent failures when used with the EDB Airman MCP component
upm-ai-model-server-beacoOOMKills due to insufficient memory limits- Updated Langflow flow changes aren’t immediately reflected in HM after republishing
- Migrations
- Supported Postgres distributions