Resolved issues
Resolved issues release notes for 24.4.x.
24.4.0-1
Updated 12/17/2024
| Issue Key | Component | Description |
|---|---|---|
| VER-97119 | UI - Management Console | In MC v11.0, the Query Profile page was working without any issues. MC v12.0 was localized and all labels were translated. However, the Query Profile page was not localized by oversight. This issue has been resolved. |
| VER-97232 | Data load / COPY | Previously, when querying certain ORC/Parquet files in a certain way, a hang would occur. This issue has now been fixed. |
| VER-97337 | Optimizer | An issue occurred, where a join spill was happening when processing the optimized delete part of the plan, which disallows Filter/Filter join distribution. Since the join inner was too large to fit in memory, an error was returned with a hint to retry with Filter/Filter enabled, even though it was already enabled in the session. This issue has been resolved; to address this, when the EnableFilterFilter hint is encountered when it's already enabled, we check if it's an optimized delete plan and in that case retry with optimized delete disabled, allowing the optimizer to choose a Filter/Filter join distribution. |
| VER-97469 | Client Drivers - VSQL | There was an issue that was preventing disabling or enabling of subcluster interior load balancing from taking effect immediately because an internal cache was not being cleared. Using ALTER DATABASE to set the EnableInteriorLoadBalancing parameter would have no impact on the result of the DESCRIBE_LOAD_BALANCE_DECISION function, which in turn would not change load balancing behavior of clients. This problem has been resolved. |
| VER-97540 | Optimizer | The view referenced in the UPDATE's target list had its subquery pruned during query rewrite, which misaligned a predicate reference to it and caused it to return a null value. This issue has been resolved; now, subquery targets are not pruned from the top-level UPDATE query's target list, and instead are only pruned FROM subqueries. |
| VER-97559 | Procedural Languages | In Vertica v.24.3 in Eon Mode, stored procedures triggered by the scheduler could not see the current subscriptions, leading to failures for certain procedures. This issue has been resolved; now, scheduled stored procedures can see the subscriptions. |
| VER-97585 | Backup/DR | If we point the --communal-storage-location at the root of a bucket, and without a trailing slash, the show-restore-points does not automatically add the trailing slash when handling the path. We fixed this issue by canonicalizing the path when we get it from the show-restore-points command. |
| VER-97763 | Execution Engine | In some cases, FCSVParser used to core dump when loading multiple files with glob patterns. This issue has been resolved. |
| VER-97867 | License | In some cases, automated database size audits used to run at a different time of the day from the one specified by GET_AUDIT_TIME. This issue has been resolved. |
24.4.0-0
Updated 10/29/2024
| Issue Key | Component | Description |
|---|---|---|
| VER-60917 | Installation Program | When installing Vertica on a remote machine, the install would hang if the /etc/motd file contained a password entry. This issue has been fixed. |
| VER-82831 | Installation Program | To prevent Vertica from running out of resources, the root user's max PID value default is changed during the Vertica install to 203816. |
| VER-84267 | Optimizer | FKPK Joins over projections with derived expressions would put PK input on the Inner side, even when it was much bigger than FK input, which resulted in worse performance in some scenarios. This issue has been resolved. |
| VER-88686 | Client Drivers - JDBC, Execution Engine | Batch update statements which run multiple query plans would keep in memory query plans from each statement in the batch until the entire batch completed. This memory bloat could cause the cluster to crash due to running out of memory. This issue has been fixed; now the query plan for each statement in the batch is freed upon completion of that statement. |
| VER-91515 | Security | install_vertica will no longer prefer the system openssl library to the one shipped with Vertica on non-fips systems. If you would like to use the system openssl library, delete the openssl libraries located at /opt/vertica/lib. |
| VER-93368 | Optimizer | Under certain circumstances, partition statistics could be used in place of full table statistics, leading to suboptimal plans. This issue has been resolved. |
| VER-94478 | AMI |
The {{install_vertica}} command fails if Transparent Huge Pages (THP) are incorrectly configured. The function that automatically enables THP has been removed. For on-premises installations, follow the steps here [https://docs.vertica.com/24.2.x/en/setup/set-up-on-premises/before-you-install/manually-configured-os-settings/] to manually configure OS settings and disable THP. If THP is not disabled, {{install_vertica}} will return an error. For installations where THP needs to remain enabled, add the parameter {{--failure-threshold NONE}} when running {{install_vertica}} to bypass this requirement and complete the installation. |
| VER-94504 | Execution Engine | An issue that caused a crash while using WITHIN GROUP () function with listagg has been fixed. |
| VER-95014 | Backup/DR | VBR no longer requires compat-openssl11 to be installed on RHEL9 systems. |
| VER-95097 | Machine Learning | There is a corner case where an orphan blob may remain in a session when the training of an ML model is cancelled. This orphan blob could cause a crash if there was an attempt to train a model with the same name on the same session. This issue has been resolved. |
| VER-95199 | Optimizer | Under certain circumstances, partition statistics could be used in place of full table statistics, leading to suboptimal plans. This issue has been resolved. |
| VER-95497 | Execution Engine | Due to a bug in the numeric division code, users would get a wrong result when evaluating the mod operator on some numeric values with large precision. This issue has been resolved. |
| VER-95651 | Kubernetes | If a database has a corrupted or outdated catalog (where some nodes do not have the latest catalog), it may not revive correctly in Kubernetes. A workaround is available to recreate nodes with an outdated catalog. |
| VER-95721 | Execution Engine | An error in expression analysis for REGEXP_SUBSTR would sometimes lead to a crash when that function was in the join condition. This error has been resolved. |
| VER-95884 | Data Export | Vertica previously supported numeric precision up to 38 digits when writing to and reading from the Parquet format. This precision limit has now been increased to 153 digits. |
| VER-95988 | Backup/DR | Server-based replication with target namespace used to fail due to namespace name case sensitivity. This issue has been resolved. |
| VER-96152 | EON | Previously, in certain cases when a cancel occurred during Vertica uploads to the communal storage, the node would crash. This issue has now been fixed. |
| VER-96243 | Catalog Engine | Qualified schema names with namespaces have some restrictions, which led to instances of no results being returned. This issue has been resolved. Note: This issue only occurred with displaying the system table. |
| VER-96866 | Security | Clients not sending packets during initialization of TLS connections will no longer cause CPU usage to spike. |
| VER-96876 | Node Management Agent | When scaling down a cluster in kubernetes, the nodes fail to delete. This issue has been fixed. |
| VER-97019 | Admin Tools, Installation Program, Installation: Server RPM/Deb | The urllib3 version is now upgraded from 1.26.18 to 1.26.19. |
| VER-97022 | Optimizer | Between v12 and v24, a previous bug fix made it so that null rows passed hash SIP filters. This led to a performance drop on queries that relied on SIPS to filter out nulls early. This has been resolved; now, SIP filters remove null rows again. |
| VER-97047 | Kubernetes | If a customer has a database with multiple local storage locations (e.g. multiple DATA paths) that don't have a common path, they cannot revive the database in K8s. This issue has been resolved by removing this limitation in K8s. |