This is the multi-page printable view of this section. Click here to print.
List of unresolved issues for this Vertica release version.
List of resolved bugs and fixes for all hotfixes in this Vertica release version.
Describes the new features introduced in this Vertica release, and provides references to detailed information in the documentation set.
Details all deprecated and removed functionality for this Vertica release version.
Vertica makes every attempt to provide you with an up-to-date list of significant known issues in each release. We will update this list as we resolve issues and as we learn of new issues.
Updated 04/01/2025
| Issue Key | Component | Description |
|---|---|---|
| VER-97978 | Installation Program | When upgrading from release 24.1.0-6 to 24.3.0-4, Vertica doesn't remove the libssl* and libcrypto* files from /opt/vertica/lib. |
| VER-93373 | Client Drivers - JDBC | The JDBC driver returns the octal values when retrieving the ARRAY (VARBINARY) data. |
| VER-78310 | Client Drivers - JDBC |
JDBC complex types return NULL arrays as empty arrays. For example, when executing this SQL statement: The array column the server returns will be: Because of the null values in the string literal, it should return this value: This is a work around to a limitation in Simba. |
| VER-78074 | Procedural Languages, UDX | Stored procedures that contain DML queries on tables with key constraints do not return a value. |
| VER-61069 | Execution Engine | In very rare circumstances, if a Vertica process crashes during shutdown, the remaining processes might hang indefinitely. |
Updated 04/28/2026
Internal release only.
Updated 04/21/2026
| Issue Key | Application Module | Description |
|---|---|---|
| 5297495 | Admin Tools | This update resolves multiple critical and high-severity security vulnerabilities, including CVE-2025-68121, CVE-2024-37298, CVE-2025-61726, CVE-2025-61728, CVE-2025-61730, CVE-2025-66418, CVE-2025-66471, CVE-2026-21441, CVE-2026-24049, and CVE-2026-26007. These issues were identified in version 23.4.0-23 and have been remediated to enhance system security. |
| 5331911 | Backup/DR |
This update optimizes the load_snapshot() operation to significantly reduce execution time. Previously, redundant CRC checks for existing storage bundles and repeated storage location verifications for Storage Containers could cause delays of up to 4 hours. This fix streamlines these processes, resulting in a more efficient and faster snapshot loading experience. |
| 5333791 | Backup/DR | Tombstone objects can have empty locations if queued after an upgrade from version 23.4 and older. For these, we now use the default storage location to avoid possible SIGSEGV. |
Updated 03/20/2026
| Issue Key | Application Module | Description |
|---|---|---|
| 4837348 | Execution Engine | Certain complex queries using internal temp relations in their plans could either error or output duplicate rows. This issue has been resolved. |
| 4866836 | Catalog Engine | Fixed a node crash issue that occurred when running ALTER TABLE commands. The crash was triggered by a VAssert error (!inParentCatalog || (inParentCatalog && parent)) within the TieredCatalog component. This issue has been resolved to prevent unexpected node failures during ALTER TABLE operations. |
| 4989561 | Execution Engine | An issue was found, where a node crash was triggered by a VAssert failure related to memory allocation, specifically 'VAssert(toCheckCleared[0] == 0 && (memSize < 16 || toCheckCleared[1] == 0)) failed'. This fix addresses the underlying memory management issue to prevent node crashes in similar scenarios. |
| 5016484 | Admin Tools |
Previously, running admintools The issue has been fixed — the help output now correctly shows: |
| 5080225 | Execution Engine | Replicated tables used to error out when performing cross namespace ILM operations. This issue has been resolved. |
| 5084591 | Backup/DR | Fixed an issue where server-based replication between two Eon Mode clusters failed due to mismatched shard boundaries after migrating from an Enterprise Edition (EE) cluster to an Eon mode cluster. Replication now functions correctly between migrated and freshly installed clusters. |
| 5086371 | Data Networking | Addressed a potential deadlock scenario where TS callback destructors were executing while holding a lock. This could lead to conflicts with other system components, such as the async Messenger notifier, which might also require the same lock. The system has been updated to ensure that TS callback destructors now execute outside of the lock, thereby preventing deadlocks and improving overall system stability. |
| 5087105 | Data Networking |
A high-severity issue where the Vertica database would hang shortly after upgrading to version 25.3.0-2, particularly when MinTLSVersion was set to 3, has been resolved. This fix, included in release VER 25.3.0-5, ensures database stability and prevents SQL queries from becoming stuck on all nodes, restoring normal operation. |
| 5093600 | Backup/DR | Int-String comparison in table expression used to fail VBR and VBR5 restore. This issue has been fixed. |
| 5105439 | UI - Management Console |
The Management Console has been updated to no longer query the EnableSSL configuration parameter, eliminating the daily error messages and false alarms. |
| 5107271 | Optimizer |
Resolved an issue where UPDATE statements, when joining with multi-level views (views referencing other views and tables across different schemas), could incorrectly trigger an 'Internal Optimizer Error (11)'. Update queries involving these complex view structures now execute as expected without error. |
| 5108701 | System: Performance | This fix involved changing how we track scan columns for the general case vs. in a database that has been resharded. In v11.1, the resharding feature added tracking of segmentation columns to all scans, even when they are not needed - the fix was to only track the segmentation columns if they are specifically needed for the query or for a reshard scenario. |
| 5143061 | Supported Platforms |
Several critical and high-severity security vulnerabilities (e.g., CVE-2025-58183, CVE-2025-58186, CVE-2025-58187 in stdlib) within the vertica-23.4.0-21.x86_64.SUSE.rpm package have been remediated. |
| 5159224 | Execution Engine | Vertica was not be able to process queries with a user defined function in an array. Now it handles this case as expected without risk of crash (in the worst case scenario). |
| 5177723 | DDL - Table |
This fix resolves an issue where temporary relation (TempRel) decoding could lead to data corruption, specifically affecting string data within recursive Common Table Expressions (CTEs). The fix ensures proper handling of character encoding and VARCHAR length constraints, preventing the occurrence of non-UTF8 data and oversized strings, thereby restoring data integrity and eliminating the need for performance-impacting workarounds. |
| 5184561 | Admin Tools | Before this release, AdminTools restart_node may find a secondary UP node to run catalog editor for getting catalog information. After this fix, AdminTools restart_node consistently finds a primary UP node to run catalog editor for getting catalog information, this ensures the getting the correct and up-to-date catalog information. |
| 5196647 | Execution Engine | An enhancement was added in version 24.1 which preserved the outer's sortedness for hash/merge right outer joins, if it was compatible. With this ticket, we saw that it is possible in a merge join for that enhancement to result in a sort violation (which can create inconsistent results with aggregate functions, as in this case). In order to maintain the sort, we added back the explicit Sort path when we're dealing with a merge join. |
| 5202439 | Execution Engine | Previously when a user using the COPY command with the delimited parser was trying to insert data for an array[varchar] column, they could trigger an internal error if their input began with back-to-back enclosing characters. Now, the server appropriately can parse multiple enclosing characters to begin an array[varchar] column, where the first enclosing character is treated as an enclosing character and the second enclosing character represents the first character in the varchar data. |
| 5203746 | Kafka Integration |
This update addresses a critical security vulnerability, CVE-2025-68161, found in org.apache.logging.log4j_log4j-core version 2.17.1. The vulnerability was identified during Prisma scans on container_minimal_arm environments. The Log4j core library has been upgraded to version 2.25.3 to mitigate this issue. |
| 5206360 | Catalog Engine |
Concurrent sequence generation in OCC lock mode can now tolerate LockTimeout setting to zero. Instead of immediate failing, they now wait for the maximum amount of time given by LockTimeout and a new knob called MinOCCLockTimeout. |
| 5211012 | EON | We used to determine the transaction log size based off a wrong counter hence truncating a corrupted transaction file more than necessary. We now use a right counter to store the size hence truncating the last transaction file correctly. |
| 5214187 | S3 |
Previously, the S3 layer could incorrectly count streaming connections, potentially leading to connection pool exhaustion during high S3 request surges. This could block S3 read operations and queries, including critical services like TxnLogSync. This fix ensures streaming connections are counted properly and limits are enforced, preventing connection starvation and improving S3 read reliability. |
| 5222230 | AMI | Queries failing due to network errors in EE's Data Target Proxy are now retryable, allowing the Vertica server to mitigate such failures with its retry mechanism. |
| 5224614 | Optimizer | Previously, nodes could be brought down by executing a MERGE query when a target value of the INSERT statement was a subquery. Now, a proper error message is displayed when this is attempted which is in line with the restrictions for using a subquery as a value for an INSERT statement outside of a MERGE query. |
| 5225289 | Kubernetes, Vcluster | Before the fix, Kubernetes operator may error out upon seeing remote user storage location when reviving an Eon mode database migrated from a 10.x Enterprise database. Now, Kubernetes operator correctly handles the remote user storage locations created in an old version enterprise database for reviving. |
| 5225681 | Optimizer |
This release addresses an 'Internal Optimizer Error (11)' that could occur when executing complex SQL queries involving Common Table Expressions (CTEs) and UNION operations. The error was observed in scenarios where the Previously Affected Versions: Vertica v25.1.0-0, v24.4.0-0, v23.4.0-14, and earlier. Workaround (prior to this fix): Users could mitigate the issue by setting the This issue is now resolved, allowing such queries to execute successfully without requiring parameter adjustments. |
| 5232013 | Node Management Agent, Vcluster | This update addresses multiple security vulnerabilities, including critical and high severity CVEs, identified in the Golang binaries shipped with Vertica 26.1.0-0 container images. These vulnerabilities previously impacted deployment in Kubernetes environments utilizing vulnerability-based admission controllers. The fix enhances the security posture of Vertica container images and ensures compliance for seamless integration and deployment. |
| 5244104 | Optimizer | In some cases, analyze statistics could encounter internal type mismatches when processing projections causing an Optimizer assert failure. The cause of these mismatches has been resolved. |
| 5256912 | Execution Engine |
Resolved a critical defect where the Vertica server (v25.2.0-3) would crash unexpectedly with a SIGSEGV error when executing queries involving the ~ ANY (Subquery) construct. This issue led to database unavailability.This fix addresses the underlying cause of the server crash, ensuring stability and preventing database panics when using ~ ANY (Subquery) in queries. |
| 5266217 | Optimizer | In some cases, union sub-queries with coercion could result in type conflicts and an internal error. This has been resolved. |
Updated 12/05/2025
| Issue Key | Application Module | Description |
|---|---|---|
| 4878505 | System: Performance |
Fix for 100% CPU Utilization with STRPOS Function: We’ve optimized the behavior of the STRPOS function to significantly improve performance in specific scenarios. What changed: Previously, all inputs to STRPOS were converted to UTF-16 to leverage the ICU (International Components for Unicode) library for string comparison. While ICU provides robust support for complex locale-aware comparisons, this conversion introduced overhead—especially for simple use cases. Now, for simple locales with 1-to-1 character mappings (such as en_US, de_DE, and other locales where each character maps directly between UTF-8 and UTF-16 without requiring normalization or multi-codepoint handling), STRPOS uses a UTF-8 based comparison. This avoids the need for conversion to UTF-16 and bypasses the ICU library entirely, resulting in much faster execution. |
| 4970328 | Security | Fixed an issue where creating access policies added extra parentheses, causing COPY_TABLE to fail with an access policy mismatch error. |
| 4990343 | Execution Engine | An issue occurred where LISTAGG queries did not consistently return NULL when no rows match the search criteria. This bug manifested when using the WITHIN GROUP clause alongside an AND NOT EXISTS subquery within the LISTAGG statement. This issue has been resolved. |
| 5004898 | EON | This release addresses an issue where cluster nodes experienced frequent crashes due to out-of-memory (OOM) errors. The problem was observed across primary and secondary subclusters, with RSS memory usage growing constantly, even on idle secondary subclusters. The fix resolves a memory leak that was affecting all hosts, regardless of hardware specifications or operating system. |
| 5011525 | Security |
New Database Parameter: Introduced a new database parameter, OAuth2JITConflictPolicy, to control OAuth2 JIT login for existing users. MERGE, the default, allows OAuth2 login for existing vertica users and grants them the appropriate auth method. This is identical to the existing behavior. If set to IGNORE then existing users require an OAuth2 client authentication granted to log in with an OAuth2 token, no automatic granting is performed. Users with psuedosuperuser are always treated as if OAuth2JITConflictPolicy is set to IGNORE, ie. they can only use authentication methods that have been explicitly granted. |
| 5040784 | Admin Tools | This fix ensures the upgrade process is successful and correctly handles the FlexTableLib library, resolving the fjsonparser incompatibility issue. |
| 5050270 | Node Management Agent, Vcluster | Critical security vulnerabilities have been identified and fixed in the Go binaries (vcluster, vcluster_server, and node_management_agent) that shipped with the Vertica database (both RPM and containers). |
Updated 09/30/2025
| Issue Key | Application Module | Description |
|---|---|---|
| 4773235 | License | An issue occurred where license audit compliance status showed an anomaly. This issue has been resolved. The GET_COMPLIANCE_STATUS() function now accurately reflects the overall compliance status, aligning with individual schema audits. |
| 4873509 | UI - Management Console | Fixed a defect where Vertica Management Console (MC) did not support SMTP configuration without authentication after Keycloak integration. The original email functionality has been restored to support SMTP configurations that do not require a username and password. |
| 4876454 | Optimizer | Fixed an issue where INSERT…SELECT statements would hang when GROUPBY PUSHDOWN was not occurring due to the presence of non-equi joins. The fix ensures that INSERT…SELECT queries with non-equi joins no longer result in indefinite hanging, restoring expected performance. |
| 4893664 | Machine Learning | Resolved an issue where the imported TensorFlow model required an unexpected number of input columns (64) instead of the expected 6 (based on tf_model_desc.json and model summary). The model now correctly predicts with the expected number of input features. |
| 4904024 | Hadoop, ResourceManager | Resolved an issue where upgrading from Vertica version 12.0.4 to 25.1 resulted in database startup failures due to timeouts. This issue occurred on replicated S3 buckets within a separate cluster running Rocky Linux 8.10. The fix addresses a signal 11 error encountered during the 'Read DataCollector' stage of the startup process, specifically related to refreshing disk space information. |
| 494350 | UI - Management Console | This fix addresses security vulnerabilities in Tomcat and Spring Security by upgrading Tomcat to version 10.1.42 and Spring Security to version 6.4.4. |
| 4950424 | Data load / COPY | Fixed a node down issue that occurred when the 'CopyFaultTolerantExpressions' parameter was enabled and FILLER was used in conjunction with specific TO_CHAR format strings. This fix prevents unexpected node failures during COPY operations using TO_CHAR with FILLER. |
| 4959200 | Cloud - Amazon | Fixed an issue where nodes would experience 'ExpiredToken' errors when reading from Parquet files in S3, leading to catalog truncation lag and requiring node restarts. The issue was caused by the system not properly refreshing AWS IAM role credentials. This fix ensures the system correctly handles token expiration and refreshes credentials as needed, preventing disruption to business operations. |
Updated 07/03/2025
| Issue Key | Application Module | Description |
|---|---|---|
| 4860219 | UI - Management Console | An issue where N2 series machines were not visible in the MC provisioning tool for GCP's us-east4-c region, despite being available in the GCP console and documented as supported, has been resolved. Users can now provision N2 series machines via MC in the specified region. |
Updated 06/14/2025
| Issue Key | Application Module | Description |
|---|---|---|
| 4811711 | Data load / COPY | In specific scenarios involving large JSON files with escape characters in string fields, the parsed string value could be missing a few characters. This issue has been resolved. |
| 4815847 | Client Drivers - VSQL | From Vertica 9.3, file names with special characters are in the UTF-8 format. If there are issues loading files with special characters in their names, ensure that the encoding, code page, and locale settings are configured to support UTF-8. |
Updated 05/22/2025
| Issue Key | Application Module | Description |
|---|---|---|
| 4534858 | Catalog Engine, Tuple Mover | A performance degradation that occurred after upgrading from 12.0.4 to 24.3.x and 24.4.0-0 has been resolved. |
| 4535429 | License | Table error_messages used to frequently contain warning message entries with a description similar to: "Skipping node information due to: Unknown node_id". This message did not indicate a real problem and thus has been removed. |
| 4634967 | Kafka Integration | There was a keystone configuration issue where the Kafka scheduler could not be created on Vertica 25.1 systems, regardless of whether they were running in FIPS mode or not. This issue has been resolved. |
| 4753286 | Third Party Tools Integration | Issues identified in the node_management_agent and vcluster components of Vertica have been resolved. |
| 4753290 | Admin Tools |
The following security vulnerabilities in the Python packages bundled with Vertica 25.2.0-1 have been resolved:
|
| 4771895, 4771898, 4772161 | UI - Management Console, Security | Security vulnerabilities within a component library using the Spring framework have been remedied. |
Updated 04/11/2025
| Issue Key | Application Module | Description |
|---|---|---|
| 4336550 | Complex Types | An issue was fixed, where Vertica was unable to load Parquet files with two-level encoded arrays, resulting in the error "Source has unsupported ARRAY type at X." This enhancement enables Vertica to properly process Parquet files with this array structure, improving compatibility with external data sources. |
| 4338772 | Scrutinize | An issue existed where Scrutinize wasn't collecting OS release information across various Linux distributions, including Debian, Ubuntu, Amazon Linux and SUSE Linux Enterprise. This issue has been resolved. |
| 4344794 | UDX | Creating user-defined extensions (UDX) in Python with the DEPENDS clause when the dependent libraries are present in Azure blob storage or Google cloud storage fails. This issue is now resolved. The value of maxNest is increased to 15 from a default value of 2. |
| 4346275 | Optimizer | An issue was discovered where, under some conditions, queries to tables with partitions that include attributes in non-table order could error on statistic lookup. This issue has been resolved. |
| 4346342 | License | An issue was causing inaccurate database size estimations due to cache invalidation issues after node subscriptions changed (node restarts or failures), primarily affecting sample caching for license auditing. This issue has been resolved. |
| 4347675 | Scrutinize | An issue was found, where 'save-system-table-as-parquet' option in Scrutinize corrupts directory names for exported system tables. This issue has been resolved by correcting the directory renaming logic, ensuring that the original directory names are preserved during the export process. Users can now reliably load the exported Parquet files without encountering directory name corruption. |
| 4347993 | Backup/DR | The backup process has been updated to correctly handle scenarios where one or more Vertica nodes are down. Previously, backups could fail if a node was unreachable, even when specifying the '--nodes' option. This fix ensures that the backup process only attempts to connect to active nodes, resolving the issue. |
| 4348000 | Execution Engine | The Vertica query parser now properly handles the use of PARTITION ROW and PARTITION LEFT JOIN in queries. Instead of returning internal errors, Vertica now rejects the syntax with an appropriate error message, similarly to how it handles other unsupported partition options like PARTITION AUTO, PARTITION BEST, and PARTITION NODES. This fix ensures that users receive clear and informative error messages when attempting to use unsupported partition options. |
| 4348485 | Scrutinize, Vcluster | System table data retrieved using vcluster scrutinize now has column headers in the retrieved files. This enhancement ensures compatibility with database loading processes and tools like the grasp loader, which rely on these headers to identify columns. |
| 4348547 | Directed Query | Directed query search could fail to match queries in cases where the predicate included constant-to-constant operations {{CONST op CONST. }}. This issue has been resolved, ensuring reliable query matching. |
| 4348750 | Optimizer | Historically, the statistics histogram would be invalidated if the row count was updated without fully reanalyzing the columns; this invalidation no longer happens. |
| 4348821 | Installation Program | An issue where installing Vertica on Amazon 2023 Linux fails and an error message is displayed has been resolved. |
| 4349067 | Database Designer Core, Refresh | Fixed a bug that caused some refreshes to fail when the ancient history mark (AHM) was updated in the background. The refresh process now has priority in cases where AHM and Refresh conflict, ensuring successful completion. |
| 4349129 | UI - Management Console | The 'DBD upload file permission error' has been resolved. Users can now successfully upload query files in the DBD tab, or via MC > New design > Manual, without encountering a 'Permission denied' error. This fix ensures that superusers, and all users, can upload files without permission restrictions. |
| 4349189 | Backup/DR | The garbage collection process was failing to clear unreferenced objects due to a 'file not found' error during the rebuild of the backup manifest. This issue has been resolved. |
| 4349838 | Optimizer | Queries utilizing WITH statements, used on multiple query levels with different target lists on each level, would sometimes return an error. The issue has been resolved. |
| 4350031 | Data load / COPY | We have extended our Parquet parser to support the legacy LIST structures. |
| 4350074 | Machine Learning | A fix has been implemented to address an issue encountered when importing XGBoost models in PMML format. Specifically, an error was triggered during the import process, resulting in a 'VIAssert' failure. This issue has been resolved to ensure successful PMML model imports. |
| 4350099 | UI - Management Console | An issue with a potential cross-site scripting vulnerability has been resolved. |
| 4350134 | Admin Tools | Admintools was unable to stop a node on Debian. The issue was due to changes in the 'ps' command utility in Debian 12's procps package, and has now been resolved. |
| 4350653 | Admin Tools, Backup/DR, S3 | IMDSv2 support for config setup at db start has been added. |
| 4391327 | Optimizer | A fix has been implemented to address an issue where the UNION operation was dropping TIMESTAMP precision. |
| 4391353, 4391363, 4391364, 4391366 | DDL - Table | Upgrading or migrating from v23.4 (or older) to v24.1 (or newer) invalidates the reference_schema_id field in the system table vs_view_tables (sets it to 0). This issue has been resolved. |
| 4392283 | Complex Types | When running an EXPLODE or UNNEST function in a locale with collation other than binary, Vertica would throw an error due to an improperly formatted OVER clause, even when the skip_partitioning parameter is set to true. This has been resolved; now, Vertica accepts the skip_partitioning parameter regardless of the current locale's collation. |
| 4393179 | Catalog Engine | A fix has been implemented to address reported slowness of ETL jobs due to locks when taking new sequences, impacting performance. |
| 4393180 | Execution Engine | Fixed an issue where executing ANALYZE_STATISTICS(' ') with a space as an argument would cause the node to go down. |
| 4403837 | Security | There was an issue where S3 bucket credentials were not masked when using the SHOW CURRENT command, potentially exposing sensitive information. This issue has been resolved, ensuring that credentials are now properly hidden, maintaining data security. |
| 4403814 | Catalog Engine | A node crash caused by GCLX being held by Manage Epochs: Advance AHM has been resolved. |
| 4405841 | Backup/DR | An issue where the REPLICATE function fails in an EON mode setup when a node in a subcluster goes down during the replication process has been resolved. This fix ensures that the REPLICATE function is more resilient to node failures in subclusters during EON mode operations. |
| 4446899 | EON | The UsePeerToPeerDataTransfer parameter can now be set at the subcluster level in Vertica. This enhancement allows users to reduce overhead from Peer-to-Peer (P2P) data transfer without disabling the feature at the database level, providing more granular control over performance optimization. |
| 4521827 | Third Party Tools Integration | |
| 4527291 | Security |
The rsync connection for Linux file systems has been upgraded from version 3.2.7 to version 3.4.0 to fix these security vulnerabilities:
|