This Innovation Release includes features and enhancements.
Release date: July 23, 2026
Release cadence
HM now follows a dual release strategy: LTS and Innovation Releases. For details on this model, see Hybrid Manager (HM) dual release strategy.
Release highlights
Improved accessibility: The HM console now offers better color contrast, full keyboard navigation, and improved screen-reader support as part of ongoing work toward WCAG 2.1 AA compliance.
New Chat Agent models: Chat Agent now supports Claude Opus 4.8, Claude Sonnet 5, Claude Fable 5, and Gemini 3.5 Flash, expanding your options for AI-assisted operations.
Refreshed Project Overview and Estate views: The Project Overview page has a more consistent, estate-like layout, and Estate cluster views got several usability fixes, including a new experimental list view for estate clusters.
Persistent Grafana dashboards: Custom dashboards you create now survive HM upgrades instead of being lost, so you no longer need to manually back them up.
More backup flexibility: Backups can now use LZ4 compression, and you can switch over to — or restore a backup onto — a replica cluster using a different storage class than the primary.
Postgres refers to all distributions supported by EDB
- PostgreSQL
- EDB Enterprise Postgres - formerly, EDB Postgres Extended Server (PGE)
- EDB Enterprise Postgres (Oracle Compatible) - formerly, EDB Postgres Advanced Server (EPAS)
Release details
Hybrid Manager (HM) platform
| Type | Description |
|---|---|
| Enhancement | Increased the default read timeout for Barman object store operations (backup, WAL archive, and WAL restore) from 60 seconds to 10 minutes. This setting is now exposed directly on the object store configuration in the cluster YAML, so you can tune it further if your storage connectivity requires it. |
| Enhancement | A replica cluster's storage class no longer has to match its primary's when performing a switchover or restoring a backup. You can now switch over to a replica cluster provisioned with a different storage class, and restore a backup to a cluster using a different storage class, as long as the target storage class supports volume snapshot restores. |
| Enhancement | Added support for LZ4 compression for base backups, in addition to the existing support for WAL archive compression. You can now set LZ4 as the data compression type when configuring backups for a cluster. |
| Enhancement | Added new edbctl operand subcommands to help you build custom operand images: annotate-extensions and show-extensions record and display a Postgres image's extensions as OCI annotations, and read-version, write-version, and copy-version manage an image's operand set version annotation. |
| Enhancement | Added support for Azure Key Vault-backed transparent data encryption (azure_kms) on AKS clusters. You can now select azure_kms as a transparent data encryption method for Postgres clusters deployed on AKS. |
| Enhancement | Added support for configuring a custom CA certificate for Image Library registry connections. You can now supply your own trusted CA bundle so that image discovery can connect to a private or internal-CA container registry over HTTPS without enabling insecure registry mode. |
| Enhancement | Added support for configuring an IP allowlist (CIDR blocks) for private load balancers in the console and edbctl. This enhancement provides a simpler alternative to manually defining AWS load balancer annotations for private networking. |
| Bug fix | Fixed an issue whereby failed and pending backups, which lack a start time, were listed in a non-deterministic order in the HM console. Backups without a start time now fall back to their creation time, then name, ensuring a stable, consistent order across list requests. |
| Bug fix | Fixed an issue whereby the sslutils package was missing from the EDB Postgres Advanced Server 18 operand image, which prevented CREATE EXTENSION sslutils from working on that version. |
| Bug fix | Fixed an issue whereby a backup that had exceeded its retention period could still be selected as a restore source, causing the restore to fail with a confusing backup does not exist error. Targeted restores now validate the backup's retention status before provisioning begins and are rejected if the backup is no longer retained. |
| Bug fix | Fixed an issue whereby creating a replica cluster used the primary cluster's Transparent Data Encryption (TDE) key configuration instead of its own. Replica clusters now use their own specified TDE key ID, allowing primary and replica clusters to reference distinct customer keys. |
| Bug fix | Fixed an issue whereby restoring a cluster from a volume snapshot backup could fail with a 404 error. Retention validation was incorrectly checking for a Barman backup even when the backup was a volume snapshot, causing the restore request to be rejected. |
| Bug fix | Fixed an issue whereby image sets larger than 10 MiB were silently truncated during decompression, causing the image set cache update to fail with a misleading YAML parsing error. The decompression size limit was raised to 32 MiB, and exceeding it now returns an explicit error instead of returning truncated data. |
| Bug fix | Fixed an issue whereby a trailing slash in the configured image discovery registry URL produced invalid image references during installation, causing failures when provisioning images. |
HM console
| Type | Description |
|---|---|
| Enhancement | Improved the Project Overview and Estate cluster views: fixed the estate grid's empty state, cluster status card resizing, and the view migrations link (now labeled View migration databases). The page also has a more consistent, estate-like layout, updated icons, smoother project switching, and a new experimental list view for estate clusters. |
| Enhancement | Improved accessibility across the console toward WCAG 2.1 AA compliance, including better color contrast, keyboard navigation and focus management, accessible names for icon-only buttons and links, corrected heading structure, restored pinch-to-zoom, and new automated accessibility testing in CI. |
| Enhancement | Updated the Superset, Metabase, and Airflow certified apps in the marketplace to newer versions, including security fixes and bug fixes from upstream projects. |
| Enhancement | Improved Chat Agent context tracking so that prompt budget usage is now tracked across tool calls within a given context window, providing better handling of output text before and after tool calls. |
| Bug fix | Fixed an issue whereby database configuration parameters that are fixed or blocked by the CNPG operator were still exposed as editable in HM, which could cause cluster deployment to fail if a user set one of these values. These parameters are no longer shown as configurable in the HM console. |
| Bug fix | Fixed an issue whereby OSS Library app deployment could fail on GKE ARM64 (N4A/Axion) clusters: the HammerDB app was missing a required ARM64 node toleration, and requesting a persistent volume smaller than 4Gi could leave the PVC stuck in a Pending state. When deploying OSS Library apps on GKE ARM64 node pools, set the persistent volume size to 4Gi or larger. |
| Bug fix | Fixed an issue whereby the pgAdmin 4 pod crashed with exit code 126 on OpenShift because the restricted-v2 security context constraint prevented the container from running as the required UID. pgAdmin 4 now deploys with the appropriate security context constraint and user ID configuration, allowing it to start successfully on OpenShift. |
| Bug fix | Fixed an issue whereby the TDE key list on the cluster provisioning page didn't filter keys based on the selected location, causing keys from other locations to appear as selectable options. |
| Deprecation | Removed the SQL Runner tab from the cluster detail page in the console. |
Observability and monitoring
| Type | Description |
|---|---|
| Enhancement | Added support for image repository rule events in the Activity Log. These events now display as Image Repository Rule instead of Unknown. |
| Enhancement | Grafana dashboards that you create are now persisted across HM upgrades. Previously, custom dashboards were stored in the Grafana container and were lost on upgrade, requiring manual backup. Grafana now uses persistent storage, so your custom dashboards survive upgrades while pre-built dashboards continue to be delivered as before. |
| Bug fix | Fixed an issue whereby the vmagent pod could fail to schedule on GKE ARM64 nodes, which prevented index recommendation automations from producing new indexes. |
| Bug fix | Fixed an issue whereby notification emails failed to send or displayed incorrectly due to a duplicate Content-Type header. Strict SMTP servers, such as AWS SES, rejected the messages outright, while lenient relays, such as SendGrid, delivered the emails with raw HTML markup instead of properly rendered content. |
AI Factory
| Type | Description |
|---|---|
| Enhancement | Extended AIDB pipeline support for externally-hosted CloudNativePG clusters. The pipelines and knowledge bases lists now display the cluster name, with a clickable link, for externally-hosted CloudNativePG clusters; previously, this column was left empty for these associations. |
| Enhancement | Added support for new chat models, including claude-opus-4-8, claude-sonnet-5, claude-fable-5, and gemini-3.5-flash, to the Chat Agent's recognized models list. claude-fable-5 requires the underlying inference platform account to have 30-day data retention enabled; accounts configured for zero data retention are rejected by the provider. |
| Platform support | This release includes AIDB 7.4.0. For details, see the AIDB release notes. |
| Change | Changed the AIDB extension setup validation in the beacon agent to report missing prerequisites instead of auto-configuring them. The agent now validates prerequisites—extension installation, role/grant config, and table ownership—giving you control over security-sensitive setup in Distributed High Availability (DHA) clusters (built on EDB Postgres Distributed, PGD). |
| Bug fix | Fixed an issue whereby a network policy blocked Langflow components from reaching PGD Proxy on ports 6432 and 6433, preventing Langflow from connecting to DHA clusters. |
| Bug fix | Added detailed AIDB pipeline error messages to help diagnose pipeline issues. |
Migrations
| Type | Description |
|---|---|
| Enhancement | Added foreign table names to the downloadable PostgreSQL migration assessment report. For each schema and foreign server, the report now lists the individual foreign table names (for example, tables backed by postgres_fdw, file_fdw, or oracle_fdw) instead of only a count, making it easier to identify foreign tables when planning a migration. |
| Bug fix | Fixed an issue whereby, for PostgreSQL-family sources, the schema finalization step generated an ALTER TABLE ... VALIDATE CONSTRAINT statement with an unquoted constraint name. Mixed-case foreign key constraint names were folded to lowercase by PostgreSQL, causing validation to fail and the migration to get stuck in the Finalizing Schema phase. |
| Bug fix | Fixed an issue in PostgreSQL-source migrations where tables with quoted, all-uppercase column names (for example, ALLCAPS_LOB) failed to replicate. The mapping pipeline now correctly preserves the uppercase format instead of converting it to lowercase, preventing missing-column errors on the target. |