New and changed in version 26.2

Support multiple JWT public keys and optional JWK endpoint

OpenText™ Analytics Database now supports additional key sources for JWT token validation in OAuth authentication records. Two new parameters have been added:

• jwt_ec_public_key: An EC (Elliptic Curve) public key, equivalent to the existing jwt_rsa_public_key parameter but for EC keys.
• jwt_jwks_url: A URL pointing to the identity provider's JWK endpoint. The database automatically retrieves signing keys from this endpoint, eliminating the need to manually update keys during identity provider key rotation.

With these changes, jwt_rsa_public_key is no longer required. At least one of jwt_rsa_public_key, jwt_ec_public_key, or jwt_jwks_url must be set for a JWT authentication record to be active. You can configure any combination of these parameters.

See OAuth authentication parameters for more information.

Init containers for VerticaDB on Kubernetes

Run pre-start initialization tasks using init containers in the VerticaDB custom resource. You can now define containers that execute sequentially before the database server starts to prepare dependencies and validate the environment. For more information, see VerticaDB custom resource definition.

VerticaRestorePointsQuery custom resource

Manage database restore points using the VerticaRestorePointsQuery custom resource. You can now save, view, and restore from restore points. You can also alter or drop archives. For more information, see VerticaRestorePointsQuery custom resource definition.

Iceberg namespace support

Support is added for Apache Iceberg namespaces, enabling namespace management through CREATE ICEBERG NAMESPACE, ALTER ICEBERG NAMESPACE, and DROP NAMESPACE statements. Iceberg namespace metadata is available in the V_CATALOG system tables.

Automatic unsegmentation for non-superprojection tables

OpenText™ Analytics Database can automatically create unsegmented projections for tables that are created without a superprojection when they are first populated using a COPY operation with small source files. This behavior improves performance for small tables by eliminating the overhead of distributed processing. For more information, see Unsegmented projections.

Export to Iceberg tables

OpenText™ Analytics Database supports exporting to Iceberg tables that are managed directly by a filesystem, AWS Glue, or Hive metastore. This behavior is controlled by the IcebergCatalog parameter in the EXPORT TO PARQUET command.

MCP Server enhancements

Query profiling

Analyze SQL query performance using LLM-powered analysis of execution plans and runtime metrics. Capture and store query profiles, visualize plan trees, and export profile data for troubleshooting and optimization. For more information, see Query profiling.

VCluster database operations and health watchdog

All tools available in the VCluster server are also available through the MCP server. Perform VCluster database operations through the MCP server, including cluster management and health monitoring. The health watchdog continuously tracks cluster performance, node availability, and system health, providing real-time insights for proactive monitoring and alerting. For more information, see Database management with VCluster UI.

VCluster UI query profile page

Import and analyze query profile data directly from the VCluster UI. Specify a target schema and optional Key ID to organize profiles, then use integrated profiling tools to investigate performance, identify bottlenecks, and optimize queries. For more information, see Query Profile.

Read-only VCluster server in Kubernetes

The VerticaDB Operator now supports deploying a dedicated, read-only VCluster Server alongside your Kubernetes-managed database using the new VClusterServer custom resource definition (CRD). The VCluster Server runs the vcluster_server service with all write operations disabled, preventing conflicts with the operator's lifecycle management. The operator automatically provisions the required Kubernetes objects—ServiceAccount, Role, RoleBinding, Deployment, and Service. Access requires httpsNMATLS to be configured on the VDB; clientServerTLS and interNodeTLS are optional. For more information, see VClusterServer custom resource definition.

NUMA multi-node cluster management with VCluster CLI

VCluster CLI now supports managing NUMA (Non-Uniform Memory Access) multi-node clusters, where multiple database nodes run on the same physical host on different client ports. New options for add_node and remove_node let you specify per-node client ports and control NUMA node assignment for each host. This works both on real NUMA hardware and in standard environments running multiple nodes on a single host. For more information, see NUMA multi-node clusters.

Unified console

Unified console is a database and data management platform that enables Vertica users to interact with and manage databases through a single interface. Vertica AI Assist is an AI assistant powered by Model Context Protocol (MCP) that enables database administrators to generate, execute, and optimize SQL queries.

Enhanced security: ECC support for SSL certificates

OpenText™ Analytics Database now supports Elliptic Curve Cryptography (ECC) for SSL certificates, offering a more modern and robust alternative to the traditional RSA algorithm. This enhancement significantly strengthens the database's SSL/TLS security posture, aligning with contemporary security best practices and providing a more secure communication environment.

Performance improvement for snapshot loading

This release addresses a performance bottleneck observed during the load_snapshot() operation, which could sometimes take up to 4 hours on certain nodes.

Key Improvements:

• Optimized Storage Container Checks: Reduced redundant checks for storage locations within Storage Containers (SCs).
• Streamlined CRC Verification: Introduced optimizations to skip unnecessary CRC checks for existing storage bundles.

These enhancements significantly improve the efficiency and reduce the duration of the load_snapshot() process, leading to faster system recovery and improved overall performance.

Improved CRC checker for communal location checks

The run_index_tool('CheckCRC', true, ''); call has been enhanced to more thoroughly inspect communal locations. Previously, the tool would skip unreadable files, potentially masking data corruption. With this improvement, the tool now accurately identifies and reports these unreadable files, along with their associated tables and projections. This allows customers to proactively detect corrupted data, facilitating recovery efforts and strengthening the database's data integrity validation mechanisms.

content/new-features/26.2/index.md