With version 9.2, Vertica has removed support for pre-join and range segmentation projections. If a table's only superprojection is one of these projection types, the projection is regarded as unsafe.
Before upgrading to 9.2 or higher, you are strongly urged to check your database for unsupported projections. If the upgrade encounters these projections, it is liable to fail. You must then revert to the previous installation.
Solution: Run the pre-upgrade script
Vertica has provided a pre-upgrade script that examines your current database and sends to standard output its analysis and recommendations. The script identifies and lists any unsupported projections. If the script finds projection buddies with different SELECT and ORDER BY clauses, it generates a deploy script. Run this script to remedy projections so they comply with system K-safety.
https://www.vertica.com/pre-upgrade-script/
For more information, see Fixing Unsafe Buddy Projections in the Vertica documentation.
The Release Notes contain the latest information on new features, changes, fixes, and known issues in Vertica 9.3.x.
They also contain information about issues resolved in:
The Premium Edition of Vertica is available for download at https://support.microfocus.com/downloads/swgrp.html.
The Community Edition of Vertica is available for download at https://www.vertica.com/download/vertica/community-edition.
The documentation is available at https://www.vertica.com/docs/9.3.x/HTML/index.htm.
Hotfixes are available to Premium Edition customers only. Each software package on the https://support.microfocus.com/downloads/swgrp.html site is labeled with its latest hotfix version.
Take a look at the Vertica 9.3.1 New Features Guide for a complete list of additions and changes introduced in this release.
An Eon Mode database can run entirely in the cloud (on AWS) or on-premise. You can now back up an on-premise database to AWS or to a Pure Storage FlashBlade appliance. You can perform full backups, full restores, and object restores from full backups.
To perform a cross-endpoint backup (or restore), you must set some additional environment variables. The vbr configuration file does not change. For more information, see Cross-Endpoint Backups in Eon Mode.
Beginning with this release, Vertica will be more rigorous in how it validates user input for configuration parameters.
For example, in past releases you could set a configuration parameter to a value with invalid trailing data. If you set parameter MaxClientSessions to 50.83 or 42plus, in both cases Vertica stripped off the invalid data.
Now, Vertica validates user input for MaxClientSessions, checking whether the value's format is valid for an unsigned 32-bit integer. Otherwise, it returns an error.
The favroparser now supports deflate compression. This new feature allows you to load Avro data files compressed with the deflate codec into a flex table. See Loading Avro Data for more information.
The Kafka integration in Vertica 9.3.1 has been tested with Kafka versions 2.2.1, 2.1, and 2.0. See Integrating with Apache Kafka for more information.
Vertica now provides the ability to cluster data using the bisecting k-means algorithm. See Clustering Data Hierarchically Using bisecting k-means for an introduction with an extended example. The following functions have been added:
EXPORT TO PARQUET previously logged information about the files it exports in the Vertica log. Now it additionally logs information in the UDX_EVENTS system table, allowing you to see information from all participating nodes in one place. See Monitoring Exports for more information about how to use this table.
You can use EXPORT TO PARQUET to export data to GCS. The requirements are similar to those for AWS S3; you must specify an authentication token in a configuration parameter, and data is exported directly to the target directory instead of being written to a temporary location first. For more information, see Exporting to S3 and GCS.
Partition clauses now support the function ZEROIFNULL. This function can check a PARTITION BY expression for null values and evaluate them to 0.
You can now perform the following operations on anchor tables with Top-K projections:
You can now use the following metafunctions on anchor tables with Top-K projections:
S3EXPORT now supports a new null_as parameter, which specifies how to export null values from the source data. If this parameter is included, S3EXPORT replaces all null values with the specified string.
Because privileges on database objects can come from several different sources such as explicit grants, roles, and inheritance, they can be difficult to monitor. To address this, the new GET_PRIVILEGES_DESCRIPTION metafunction has been introduced. It provides a unified view of their effective privileges across all sources on a specified database object.
The GET_METADATA function inspects a Parquet file and returns its metadata, including columns, row groups, and sizes. You can use this information to help you define external tables and to confirm the results of exports from Vertica.
The INFER_EXTERNAL_TABLE_DDL function inspects a Parquet file and returns a starting point for the definition of an external table. This function can be especially helpful for tables with many columns of mostly primitive types. For some types, the function cannot infer the data type and you must edit the output.
You can now comment on table columns with COMMENT ON COLUMN. For more information, see COMMENT ON TABLE COLUMN.
You can now trust more than one CA with the SSLCA parameter. For more information, see Security Parameters.
The LDAP_LINK_DRYRUN family of metafunctions allows you to tweak your LDAP Link settings before syncing with Vertica. Each dry run metafunction takes LDAP Link parameters as arguments and tests a separate part of LDAP Link:
These metafunctions are meant to be used and tested in succession, and their arguments are cumulative. That is, the parameters you use for configuring LDAP_LINK_DRYRUN_CONNECT are used for LDAP_LINK_DRYRUN_SEARCH, and the arguments for those functions are used for LDAP_LINK_DRYRUN_SYNC.
For detailed instructions on using these metafunctions, see Configuring LDAP Link with Dry Runs.
New parameters have been added for configuring LDAP Link.
See LDAP Link Parameters for more information.
You can now call ANALYZE_STATISTICS on global temporary tables. As with local temporary tables, you can only obtain statistics on a global temporary table that is created with the option ON COMMIT PRESERVE ROWS. Otherwise, Vertica deletes table content on committing the current transaction, so no table data is available for analysis.
When working with Parquet files containing complex types, you can now define tables using arrays and maps of primitive types, allowing you to read a larger range of existing data.
You can query arrays, use them in joins and other operations, and use Array Functions on both array columns and literals. See Reading Arrays and the ARRAY data type for more information.
You can define maps, which allows you to read files containing them, but you cannot query them. See Reading Maps and the MAP data type for more information.
Arrays and maps may contain only primitive types. They cannot contain rows, arrays, or maps.
The new MERGEOUT_PROFILES system table returns information about automatic mergeout operations, making them easier to monitor and troubleshoot.
System table LOAD_SOURCES now includes four new columns that return how much time (in microseconds) is consumed by various user-defined load functions:
The LDAP_LINK_DRYRUN_EVENTS system table contains the results from the new LDAP dry run metafunctions.
The new UDX_EVENTS system table records events logged from user-defined functions, if they logged any. See Logging in Extending Vertica for information about how to log events in UDxs that you write.
The preconfigured tokenizer StringTokenizerDelim has been optimized. It has also seen a slight change in behavior: an empty OVER() clause returns a single column for the tokenized values. Previously, the function would return an additional column for the original input. To use the old behavior, include a PARTITION BY clause in OVER().
For more information on this change and other preconfigured tokenizers, see Preconfigured Tokenizers.
Previously, create_db would rollback the entire operation on failure, deleting log files and making it difficult to diagnose problems. The following changes have been made to modify this behavior.
This section describes the two phases Vertica follows to retire Vertica functionality:
The following Vertica functionality was deprecated and will be retired in future versions:
No deprecated functionality
The following functionality is no longer accessible as of the releases noted:
For more information see Deprecated and Retired Functionality in the Vertica documentation.
This hotfix addresses the issues below.
|
Issue |
Component |
Description |
| VER-42720 | Execution Engine | Vertica occasionally generated misleading error messages in certain distributed query executions. This issue has been resolved. |
| VER-63714 | Cloud - Google | Previously, exporting parquet files to Google Cloud Storage caused an error. This is now fixed. |
| VER-66216 | Admin Tools | Passwords are now validated when creating databases. |
| VER-66903 | Kafka Integration | Users can now use kafka_conf to override the kafka ssl setting to allow SASL+SSL authentication to work. |
| VER-67029 | Client Drivers - JDBC | The maxconnection limitation check no longer fails after running a distribute query. |
| VER-67201 | Client Drivers - ODBC | Previously, the names of files used in copy local operations were not correctly represented within the Windows version of the ODBC/OLEDB drivers; as a result, non-ASCII characters could not be used in these filenames. The same problem existed for the names of rejection and exception filenames specified in a copy local operation. This problem has now been corrected, and non-ASCII characters can be used in all three types of files. |
| VER-67497 | Client Drivers - VSQL | Previously, the VSQL options -B (connection backup server/port) and -k/-K (Kerberos service and host name) were incompatible, and the latter would be ignored. Now, these options can be set together. |
| VER-67511 | Tuple Mover | Configuration parameter MergeOutInterval specifies in seconds (by default 600) how often the Tuple Mover checks the mergeout request queue for pending requests. If the queue is empty, the Tuple Mover processes storage location move requests. In previous releases, Tuple Mover threads that were triggered by DML operations ignored the MergeOutInterval parameter and processed all pending requests, including storage location move requests. This sometimes resulted in frequent checks for storage location move requests, which could adversely affect performance. To resolve this issue, Tuple Mover threads that are spawned by DML operations are now subject to MergeOutInterval. |
| VER-67701 | LocalPlanner | Vertica occasionally failed when running COUNT DISTINCT queries with predicates on lead columns in the projection sort order. This issue has been resolved. |
| VER-67964 | Execution Engine | In rare circumstances, if a cluster's partitioning was short-lived due to network problems, then the whole cluster may have gone down. This issue has been fixed. |
| VER-67966 | Backup/DR | If a user performed a copycluster task and a swap partition task concurrently, the data that participated in the swap partition could end up missing on the target cluster. This issue has been resolved. |
| VER-67980 | Client Drivers - ODBC | Error messages containing Japanese column names now display correctly. |
| VER-68414 | SAL | Queries could fail when Vertica 9.2 tried to read a ROS file format from Vertica 6.0 or earlier. Vertica now properly handles files created in this format. |
| VER-68549 | UI - Management Console | A problem occurred when an MC Admin user chose to bind to LDAP anonymously, where the setting did not take effect despite clicking 'Apply' and 'Done'. This problem has now been fixed. |
| VER-68633 | Security | To ensure that username case is consistent between SESSIONS and other system tables, the SESSIONS system tables are now populated with the username value from the catalog. |
| VER-68644 | Tuple Mover | In previous releases, Vertica considered unsafe projections when it tried to advance AHM. Now, Vertica ignores unsafe projections when it determines the AHM.
|
| VER-68720 | UDX | GRANT and REVOKE statements now support ALTER and DROP privileges for user-defined functions. |
| VER-68726 | Kafka Integration | Vertica no longer fails when it encounters improper Kafka SSL key/certificate setup error information on RedHat Linux. |
| VER-68790 | Backup/DR | If a user deleted an object during a backup on EON mode, the backup could potentially fail. This issue has been resolved. |
| VER-68853 | Subscriptions | A failure in one EON Mode subcluster no longer has the ability to affect another subcluster. |
| VER-68898 | Data Removal - Delete, Purge, Partitioning | DROP_PARTITIONS requires that the range of partition keys to drop exclusively occupy the same ROS container. If the target ROS container contains partition keys that are outside the specified range of partition keys, the force-split argument must be set to true, otherwise the function returns with an error.
In rare cases, DROP_PARTITIONS executes at the same time as a mergeout operation on the same ROS container. When this happens, the function cannot execute the force-split operation and returns with an error. This error message has been modified to better identify the source of the problem, and also provides a hint to try calling DROP_PARTITIONS again. |
| VER-69184 | Execution Engine | In cases where COUNT (DISTINCT) and aggregates such as AVG were involved in a query of a numeric datatype input, the GroupGeneratorHashing Step was causing a memory conflict when the size of the input datatype (numeric) was greater than the size of output datatype (float for average), producing incorrect results. This issue has been fixed. |
| VER-69217 | Data load / COPY | Sending excessively long inputs to a NUMERIC column in a COPY statement could cause Vertica to crash. This issue has been fixed. |
| VER-69229 | Execution Engine | In some cases, queries failed to sort UUIDs correctly if the ORDER BY clause did not specify the UUID column first. This problem has been resolved. |
| VER-69292 | Execution Engine, UDX, Vertica Log Text Search | Previously, the v_txtindex.StringTokenizerDelim UDx made copies of the input for each token, which could lead to memory issues. This function no longer makes copies of the original input and now returns one column containing the tokens.
To produce the old behavior that maps the tokens to the original input, include a "PARTITION BY <input_col>" clause inside the OVER() clause of the UDx call. |
| VER-69383 | Optimizer - Statistics and Histogram | ANALYZE_ROW_COUNT can no longer change STATISTICS_TYPE from FULL to ROWCOUNT. |
| VER-69431 | Data Export | S3EXPORT now supports a null_as parameter that specifies how to export null values from the source data. If this parameter is included, S3EXPORT replaces all null values with the specified string. |
| VER-69486 | Security | Previously, LDAP Link required a global catalog lock during the entire operation. This would sometimes cause the database to fail if the LDAP server took too long to respond. Now, LDAP Link only requires a global catalog lock when it maps LDAP entries to Vertica users and roles. |
| VER-69489 | Data load / COPY | COPY statements could automatically generate rejected data file names that were too long for Vertica's file system, causing the COPY to error out. Vertica will now truncate these file names to prevent this from happening. |
| VER-69535 | Data load / COPY | A bug in tracking disk usage of rejected data tables would rarely cause Vertica to crash. This issue has been fixed. |
| VER-69606 | Catalog Engine | ETL queries no longer fail with the error "Active subscriptions changed during query planning" if you run the query while running REBALANCE_SHARDS on another subcluster. |
| VER-69702 | Communications/messaging | Queries performed against a virtual table no longer display duplicate error messages if performed while one or more nodes is down. |
| VER-70487 | Admin Tools | Starting a subcluster in an EON Mode database no longer writes duplicate entries to the admintools.log file. |
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.
|
Issue |
Component |
Description |
| VER-41895 | Admin Tools |
On some systems, admintools fails to parse output while running SSH commands on hosts in the cluster. Workaround: If the admintools operation needs to run on just one node, users can avoid this issue by using SSH to connect to the target node and running the admintools command on that node directly. |
| VER-48020 | Hadoop | Canceling a query that involves loading data from ORC or Parquet files can be slow if there are too many files involved in the load. |
| VER-48041 | Admin Tools | On some systems, occasionally admintools cannot parse the output it sees while running SSH commands on other hosts in the cluster. The issue is typically transient. |
| VER-60409 | AP-Advanced, Optimizer |
APPLY_SVD and APPLY_PCA functions are implemented as User-Defined Transform Functions (UDTFs). A query containing a UDTF that takes many input columns (e.g. more than 1000) and also outputs many columns may fail with error "Request size too big" due to additional memory requirement in parsing. Workaround: Increase configuration parameter MaxParsedQuerySizeMB or reduce the number of output columns. For the cases of APPLY_SVD and APPLY_PCA, you can limit the number of output columns either by setting the parameter num_components or by setting the parameter cutoff. In practice, cutoff=0.9 is usually enough. Please note that If you increase MaxParsedQuerySizeMB to a larger value, for example 4096, each query you run may use 4 GB of memory during parsing. This means running multiple queries at the same time could cause out-of-memory (OOM) if your total memory is limited. Please refer to Vertica documentation for more information about MaxParsedQuerySizeMB. |
| VER-61069 | Execution Engine |
In very rare circumstances, if a Vertica process crashes during shutdown, the remaining processes might hang indefinitely. Workaround: Halt the remaining processes using admin tools. |
| VER-61420 | Data Removal - Delete, Purge, Partitioning | Partition operations, such as move_partitions_to_table, must split storage containers that have both partitions that match the operations and partitions that do not. Version 9.1 introduced an inefficiency whereby such a split may separate a storage container into one more storage container than necessary. Note that any extraneous containers created this way will eventually be merged by the TM. |
| VER-62061 | Catalog Sync and Revive | If you revive a database on a one-node cluster, Vertica does not warn users against reviving the same database elsewhere. This can result in two instances of the same database running concurrently on different hosts, which can corrupt the database catalog. |
| VER-62983 | Hadoop | When hcatalogconnector schemas are created with custom_partitions enabled, poor performance has been observed when there are many (500+) partitions. By default, custom_partitions are disabled on hcatalogconnector schemas. |
| VER-64352 | SDK |
Under special circumstances, a sequence of statements to create and drop Python libraries fails. For example, the following session statements attempt to create Python libraries lib1 and lib2: CREATE LIBRARY lib1 AS :lib1_path DEPENDS :numpy_path LANGUAGE 'Python'; DROP LIBRARY lib1; CREATE LIBRARY lib2 AS :lib2_path DEPENDS :sklearn_path LANGUAGE 'Python'; Here, lib1 is a Python library that imports module numpy, while lib2 imports module sklearn, which relies on numpy. After lib1 is dropped, the CREATE LIBRARY statement for lib2 fails. Workaround: One of the following:
|
| VER-67228 | AMI, License | An Eon Mode database created in one hourly listing can only be revived into another hourly listing with the same operating system. For example, an Eon Mode database created in the Red Hat hour listing cannot be shutdown then revived into a cluster created using the Amazon Linux listing. |
| VER-70139 | Metadata Tables Security |
The LDAP Link dry run metafunctions do not yet have full support for the LDAPLinkStartTLS and LDAPLinkTLSReqCert parameters. Workaround: When using the LDAP Link dry run metafunctions, pass "1" and "allow" for LDAPLinkStartTLS and LDAPLinkTLSReqCert arguments, respectively |
| VER-70238 | Backup/DR, Security |
A vbr restore operation can fail if your database is configured for SSL and has a server.key and a server.crt file present in the catalog directory. Workaround: Move your SSL authentication related files (server.crt, server.csr and server.key) from the catalog directory before performing the restore. |
| VER-70468 | Nimbus |
Creating a load balance group for a subcluster with a ROUNDROBIN load balance policy only has an effect if you set the global load balancing policy to ROUNDROBIN as well. This is the case, even though the LOAD_BALANCE_GROUPS system table shows the group's policy as ROUNDROBIN. Workaround: Set the global load balancing policy using the SET_LOAD_BALANCE_POLICY function: SELECT set_load_balance_policy('roundrobin'); |
| VER-70549 | Spread |
When quickly stopping and then starting a different database, the new database may fail to start after attempting to connect to Spread processes associated with the stopped database. Workaround: After stopping a database, ensure that all old Spread processes have stopped on the affected nodes before starting a different database. This typically takes no more than one minute, but may vary under certain circumstances. |
Take a look at the Vertica 9.3 New Features Guide for a complete list of additions and changes introduced in this release.
Vertica has added two Amazon Web Service (AWS) instance types to the list of supported types available in MC:
| Optimization | Type | Supports EBS Storage | Supports Ephemeral Storage |
|---|---|---|---|
|
Computing |
c5.xlarge c5.large |
Yes Yes |
No No |
Previously, Vertica meta-functions that exported DDL, such as EXPORT_TABLES, exported all table constraints as ALTER statements, whether they were part of the original CREATE statement, or added later with ALTER TABLE. This was problematic for global temporary tables, which cannot be modified with new constraints other than foreign keys. Thus, the DDL that was exported for a temporary table with constraints could not be used to reproduce that table. This issue has been resolved: Vertica now exports table constraints (except foreign keys) as part of the table's CREATE statement.
Columns in Parquet and ORC files can contain complex data types. One complex data type is the struct, which stores (typed) property-value pairs. Vertica previously supported reading structs only as expanded columns. In addition, you can now preserve the original structure by reading a struct as a single column. For more information, see Reading Structs as Inline Types.
Subclusters help you isolate workloads to a smaller group of nodes in your database cluster. Queries only run on nodes within the subcluster that contains the initiator node. In previous versions of Vertica, you defined subclusters using the fault groups feature. Starting in 9.3.0, subclusters are their own unique feature and have become more useful.
All nodes in the database must belong to a subcluster. When you upgrade an Eon Mode database from a previous version to 9.3.0, Vertica converts any existing fault groups to subclusters. If there are nodes in the database that are not part of a fault group, Vertica creates a default subcluster and adds these nodes to it. When you create a new Vertica database, Vertica creates a default subcluster and adds the initial group of nodes to it. When adding a node to the database, Vertica adds the node to a default subcluster unless you specify a subcluster.
See Subclusters for more information.
When you upgrade an Eon Mode database to version 9.3.0, Vertica converts all of the fault groups in the database to subclusters. Any nodes in the default groups are automatically assigned to the converted subclusters. Any nodes that are not part of a fauilt group are assigned to a default subcluster that Vertica creates.
Subclusters come in two types:
Nodes in Eon Mode databases are also either primary or secondary, based on the type of subcluster that contains them. See Subcluster Types and Elastic Scaling for more information on primary and secondary nodes and subclusters.
Connection load balancing is now aware of subclusters. You can define connection load balancing groups based on subclusters. See About Connection Load Balancing Policies for more information.
When Vertica upgrades an Eon Mode database to version 9.3.0 or beyond, it does not convert load balancing groups based on fault groups into groups based on the converted subclusters. You must redefine these load balance groups to be based on the newly-created subclusters yourself.
The ADMIN_TOOLS_EXE command line interface has new tools to manipulate subclusters:
db_add_subcluster tool adds new subclusters.db_remove_subcluster tool removes subclustersrestart_subcluster tool starts subclusters.stop_subcluster tool stops subclusters.In addition, the add_node tool has a new --subcluster argument that lets you select the subcluster that Vertica adds the new node or nodes to.
For more information on the option, see Writing Administration Tools Scripts.
There are several new system tables, as well as changes to existing tables for subclusters:
Before a newly-added node begins processing queries, it warms its depot by fetching data into it based on what other nodes in the subcluster have in their depots. You can now choose to cancel depot warming entirely, or have the node process queries while it continues to warm its depot. See Canceling or Backgrounding Depot Warming for more information.
The PLAN_ID columns in the DEPOT_FETCH_QUEUE and DEPOT_FETCHES system tables have been replaced with the TRANSACTION_ID column. This change makes it easier to determine the transaction that caused a node to fetch a file.
Queries in Vertica mode now support the /*+DEPOT_FETCH*/ hint, which specifies whether to fetch data from communal storage when the depot does not have the queried data.
Vertica lets you set a size for the depot. This size is set to 60% by default. Vertica also uses the filesystem containing the depot for other purposes such as temporary storage while loading data. To make sure there is enough space on this filesystem for these other needs, Vertica now limits you to setting aside 80% of the filesystem space for the depot. If you attempt to allocate more than 80% of the filesystem to the depot, Vertica returns an error.
Vertica function S3EXPORT now supports the following parameters:
compression can specify the bzip2 filter to compress exported data. enclosed_by specifies the character used to enclose string and date/time data.escaped_by specifies the character used to escape values in exported data.The CLEAR_FETCH_QUEUE function now accepts an optional transaction ID parameter. Supplying this parameter limits the clearing of the fetch queue to just those entries for the transaction.
Vertica version 9.3.0 has been tested with the following versions of Apache Kafka:
Vertica may work with other versions of Kafka. See Vertica Integration for Apache Kafka for more information.
Vertica now uses version 0.11.6 of the rdkafka library to communicate with Kafka. This change could affect you if you directly set Kafka library options. See Directly Setting Kafka Library Options for more information.
Memory allocation for EXPORT TO PARQUET is now part of the Vertica resource pool instead of being separately managed, reducing memory-allocation errors from large exports.
When you rename a table with ALTER TABLE or copy an existing one with CREATE TABLE LIKE…INCLUDING PROJECTIONS, Vertica propagates the new table name to its projections. For details, see Projection Naming.
You can now run DML operations on tables with live aggregate projections. For more details, see Live Aggregate Projections.
Vertica now uses Spread 5.
Vertica now supports Pure Storage Flashblade storage for Eon Mode on premise.
This section describes the two phases Vertica follows to retire Vertica functionality:
The following Vertica functionality was deprecated and will be retired in future versions:
| Release | Functionality | Notes |
|---|---|---|
|
9.3 |
7.2_upgrade vbr task |
This task remains available in earlier versions. |
| 9.3 | FIPS | FIPS is temporarily unsupported due to incompatibility with OpenSSL OpenSSL 1.1x. FIPS is still available in Vertica 9.2.x, and FIPS will be reinstated in a future release. |
The following functionality is no longer accessible as of the releases noted:
| Release | Functionality | Notes |
|---|---|---|
| 9.3 |
|
|
| 9.3 | Configuration parameter ReuseDataConnections |
For more information see Deprecated and Retired Functionality in the Vertica documentation.
This hotfix addresses the issues below.
|
Issue |
Component |
Description |
| VER-69250 | UDX | GRANT and REVOKE statements now support ALTER and DROP privileges for user-defined functions. |
| VER-69666 | Backup/DR | If user performed a copycluster task and a swap partition task concurrently, the data that participated in the swap partition could end up missing on the target cluster. This issue has been resolved. |
| VER-69783 | Execution Engine | In cases where COUNT (DISTINCT) and aggregates such as AVG were involved in a query of a numeric datatype input, the GroupGeneratorHashing Step was causing a memory conflict when the size of the input datatype (numeric) was greater than the size of output datatype (float for average), producing incorrect results. This issue has been fixed. |
| VER-69787 | Cloud - Google | Previously, exporting parquet files to Google Cloud Storage caused an error. This is now fixed. |
| VER-70085 | Backup/DR | If a user deleted an object during a backup on Eon mode, the backup could potentially fail. This issue has been resolved. |
| VER-70144 | Communications/ messaging |
Queries performed against a virtual table no longer display duplicate error messages if performed while one or more nodes is down. |
| VER-70148 | Client Drivers - VSQL | Previously, the VSQL options -B (connection backup server/port) and -k/-K (Kerberos service and host name) were incompatible, and the latter would be ignored. Now, these options can be set together. |
| VER-70150 | Catalog Engine | ETL queries no longer fail with the error "Active subscriptions changed during query planning" if you run the query while running REBALANCE_SHARDS on another subcluster. |
| VER-70151 | Subscriptions | A failure in one Eon Mode subcluster no longer has the ability to affect another subcluster. |
| VER-70235 | Data Removal - Delete, Purge, Partitioning |
DROP_PARTITIONS requires that the range of partition keys to drop exclusively occupy the same ROS container. If the target ROS container contains partition keys that are outside the specified range of partition keys, the force-split argument must be set to true, otherwise the function returns with an error. In rare cases, DROP_PARTITIONS executes at the same time as a mergeout operation on the same ROS container. When this happens, the function cannot execute the force-split operation and returns with an error. This error message has been modified to better identify the source of the problem, and also provides a hint to try calling DROP_PARTITIONS again. |
| VER-70460 | AP-Advanced | In rare situations, certain User-Defined Aggregate functions would cause a query to return an error "Data Area overflow". This issue has been fixed. |
| VER-70466 | Admin Tools | Starting a subcluster in an Eon Mode database no longer writes duplicate entries to the admintools.log file. |
This hotfix addresses the issues below.
|
Issue |
Component |
Description |
| VER-69388 | SAL | Queries could fail when Vertica 9.2 tried to read a ROS file format from Vertica 6.0 or earlier. Vertica now properly handles files created in this format. |
This release addresses the issues below.
|
Issue |
Component |
Description |
|
VER-67087 |
Admin Tools |
Sometimes during database revive, admintools treated s3 and hdfs user storage locations as local filesystem paths. This led to errors during revive. This issue has been resolved. |
|
VER-68531 |
Admin Tools |
Previously, environment variables used by admintools during SSH operations were set incorrectly on remote hosts. This issue has been fixed. |
|
VER-64171 |
Backup/DR |
If the hardlink failed during a hardlink backup, vbr switched to copying data instead of failing the backup and threw an error. This issue has been resolved. |
|
VER-66956 |
Backup/DR |
Dropping the user that owns an object involved in a replicate or restore operation during that replicate or restore operation can cause the nodes involved in the operation to fail. |
|
VER-62334 |
Catalog Engine |
Previously, if a DROP statement on an object failed and was rolled back, Vertica would generate a NOTICE for each dependent object. This was problematic in cases where the DROP operation had a large number of dependencies. This issue has been resolved: Vertica now generates up to 10 messages for dependent objects, and then displays the total number of dependencies. |
|
VER-67734 |
Catalog Engine |
In Eon mode, queries sometimes failed if they were submitted at the same time as a new node was added to the cluster. This issue has been resolved. |
|
VER-68188 |
Catalog Engine |
Exporting a catalog on a view that references itself could cause Vertica to fail. This issue has been resolved. |
|
VER-68603 |
Catalog Sync and Revive |
Database startup occasionally failed when catalog truncation failed because the disk was full. This issue has been resolved. |
|
VER-68494 |
Client Drivers - Misc |
Some kinds of node failures did not reset the list of available nodes for connection load balancing. These failures now update the available node list. |
|
VER-67342 |
Data Removal - Delete, Purge, Partitioning Client Drivers - ODBC |
Queries use the query epoch (current epoch -1) to determine which storage containers to use. A query epoch is set when the query is launched, and is compared with storage container epochs during the local planning stage. Occasionally, a swap partition operation was committed in the interval between query launch and local planning. In this case, the epoch of the post-commit storage containers was higher than the query epoch, so the query could not use them, and it returned empty results. This problem was resolved by making SWAP_PARTITIONS_BETWEEN_TABLES atomic. |
|
VER-66546 |
Cloud - Amazon |
The verticad script restarts the Vertica process on a node when the operating system starts. On Amazon Linux, the script sometimes was unable to detect the operating system and returned an error. This issue has been resolved. |
|
VER-53370 |
DDL - Projection |
Previously, projections of a renamed table retained their original names, which were often derived from the table's previous name. This issue has been resolved: now, when you rename a table, all projections whose names are prefixed by the anchor table name are renamed with the new table name. |
|
VER-66882 |
DDL - Table |
The database statistics tool SELECT ANALYZE_STATISTICS no longer acquires a GCL-X lock when running against local temp tables. |
|
VER-67105 |
S3 Data Export |
In previous releases, s3export could not export files in CSV format. The function now supports two new parameters, enclosed_by and escape_as, that enable exporting files in CSV format. |
|
VER-68619 |
Hadoop Data Export |
Date Columns now includes the Parquet Logical type, enabling other tools to recognize this column as a Date type. |
|
VER-66272 |
Data Removal - Delete, Purge, Partitioning |
Queries use the query epoch (current epoch -1) to determine which storage containers to use. A query epoch is set when the query is launched, and is compared with storage container epochs during the local planning stage. Occasionally, a swap partition operation was committed in the interval between query launch and local planning. In this case, the epoch of the post-commit storage containers was higher than the query epoch, so the query could not use them, and it returned empty results. This problem was resolved by making SWAP_PARTITIONS_BETWEEN_TABLES atomic. |
|
VER-65427 |
Data load / COPY |
Vertica could crash if its client disconnected during a COPY LOCAL with REJECTED DATA. This issue has been fixed. |
|
VER-65659 |
Data load / COPY |
Occasionally, a copy or external table query could crash a node. This has been fixed. |
|
VER-68033 |
Data load / COPY |
A bug in the Apache Parquet C++ library sometimes caused Vertica to fail when reading Parquet files with large Varchar statistics. This issue has been fixed. |
|
VER-53981 |
Database Designer Core |
In some cases, DESIGNER_DESIGN_PROJECTION_ENCODINGS mistakenly removed comments from the target projections. This issue has been resolved. |
|
VER-62628 |
Execution Engine |
If a subquery that is used as an expression returns more than one row, Vertica returns with an error. In past releases, Vertica used this error message: ERROR 4840: Subquery used as an expression returned more than one row In cases where the query contained multiple subqueries, this message did not help users identify the source of the problem. With this release, the error message now provides the Join label and localplan_id. For example: ERROR 4840: Subquery used as an expression returned more than one row DETAIL: Error occurred in Join [(public.t1 x public.t2) using t1_super and subquery (PATH ID: 1)] with localplan_id=[4] |
|
VER-66224 |
Execution Engine |
On rare occasions, a query that exceeded its runtime cap automatically restarted instead of reporting the timeout error. This issue has been fixed. |
|
VER-67069 |
Execution Engine |
Some queries with complex predicates ignored cancel attempts, manual or via the runtime cap, and continued to run for a long time. Cancel attempts themselves also caused the query to run longer than it would have otherwise. This issue has been fixed. |
|
VER-67102 |
Execution Engine |
Users were unable to cancel meta-function RUN_INDEX_TOOL. This problem has been resolved. |
|
VER-69066 |
Execution Engine |
In some cases, queries failed to sort UUIDs correctly if the ORDER BY clause did not specify the UUID column first. This problem has been resolved. |
|
VER-64680 |
Kafka Integration |
Previously, the version of the kafkacat utility distributed with Vertica had an issue that prevented it from working when TLS/SSL encryption was enabled. This issue has been corrected, and the version of kafkacat bundled with Vertica can now make TLS/SSL encrypted connections. |
|
VER-68192 |
License |
Upgrading 8.1.1-x databases with Autopass license installed to 9.0 or later can lead to license tampering startup issues. This problem has been fixed. |
|
VER-67573 |
Nimbus |
In the past, DDL transactions remained open until all pending file deletions were complete. In Eon mode, this dependency could cause significant delays. This issue has been resolved: now, DDL transactions can complete while file deletions continue to execute in the background. |
|
VER-26260 |
Optimizer |
Vertica can now optimize queries on system tables where the queried columns are guaranteed to contain unique values. In this case, the optimizer prunes away unnecessary joins on columns that are not queried. |
|
VER-66423 |
Optimizer |
The Optimizer's is better able to derive transitive selection predicates from subqueries. |
|
VER-66902 |
Optimizer |
EXPORT TO VERTICA returned an error if the table to export was a flattened table that already existed in the source and target databases. This issue has been resolved. |
|
VER-66933 |
Optimizer |
Previously, export operations exported all table constraints as ALTER statements, whether they were part of the original CREATE statement, or added later with ALTER TABLE. This was problematic for global temporary tables, which cannot be modified with any constraints other than foreign keys; attempts to add constraints on a temporary table with ALTER TABLE return with an error. Thus, the DDL that was exported for a temporary table with constraints could not be used to reproduce that table. This problem has been resolved: the DDL for constraints (except foreign keys) is now exported as part of the table's CREATE statement. |
|
VER-66968 |
Optimizer |
Queries that perform an inner join and group the results now return consistent results. |
|
VER-67138 |
Optimizer |
When flattening subqueries, Vertica could sometimes move subqueries to the ON clause which is not supported. This issue has been resolved. |
|
VER-67443 |
Optimizer |
ALTER TABLE...ALTER CONSTRAINT returned an error when a node was down. This issue was resolved: you can now enable enable or disable table constraints when nodes are down. |
|
VER-67740 |
Optimizer |
Vertica sometimes crashed when certain combinations of analytic functions were applied to the output of a merge join. This issue has been fixed. |
|
VER-67908 |
Optimizer |
Prepared statements do not support WITH clause materialization. Previously, Vertica threw an error when it tried to materialize a WITH clause for prepared statement queries. Now, Vertica throws a warning and processes the WITH clause without materializing it. |
|
VER-68306 |
Optimizer |
Queries with multiple analytic functions over complex expression arguments and different PARTITION BY/ORDER BY column sets would sometimes produce incorrect and inconsistent results between Enterprise and Eon. This issue has been fixed. |
|
VER-68379 |
Optimizer |
Calls to REFRESH_COLUMNS can now be monitored in the new "dc_refresh_columns" table, which logs, for every call to REFRESH_COLUMNS, the time of the call, the name of the table, the refreshed columns, the mode used, the minimum and maximum key values, and the epoch. |
|
VER-68594 |
Optimizer |
Queries sorting on subquery outputs sometimes produced inconsistent results. This issue has been fixed. |
|
VER-68828 |
Optimizer |
When a session's transaction isolation mode was set to serializable, MERGE statements sometimes returned with the error message 'Can't run historical queries at epochs prior to the Ancient History Mark'. This issue has been resolved. |
|
VER-49136 |
Optimizer Optimizer - Projection Chooser |
TRUNCATE TABLE caused all existing table- and partition-level statistics to be dropped. This issue has been resolved. |
|
VER-68826 |
Optimizer - Statistics and Histogram |
If you partitioned a table by date and used EXPORT_STATISTICS_PARTITION to export results to a file, the function wrote empty content to the target file. This issue has been resolved. |
|
VER-67647 |
Build and Release Performance tests |
The ST_IsValid() geospatial function, when used with the geography data type, has seen a slight performance degradation of around 10%. |
|
VER-66648 |
QA |
When a query returned many rows with wide columns, Vertica threw the following error message: "ERROR 8617: Request size too big." This message incorrectly suggested that query output consumed excessive memory. This issue has been fixed. |
|
VER-64783 |
Recovery |
When applying partition events during recovery, Vertica determined that recovery was complete only after all table projections were recovered. This rule did not differentiate between regular and live aggregate projections of the same table, which typically recovered in different stages. If recovery was interrupted after all regular projections of a table were recovered but before recovery of its live aggregate projections was complete, Vertical returned an error. This problem has been resolved: Vertica now determines that recovery is complete when all regular projections are recovered, and disregards the recovery status of live aggregate projections. |
|
VER-68504 |
Execution Engine S3 |
If you refreshed a large projection in Eon mode and the refresh operation used temp space on S3, the refresh operation occasionally caused the node to crash. This issue has been resolved. |
|
VER-60036 |
SDK |
All C++ UDXs require c++11 standard now. That can be done by setting -std=c++11 flag at compilation. |
|
VER-61780 |
Scrutinize |
Scrutinize previously generated a UnicodeEncodeError if the system locale was set to a language with non-ASCII characters. This issue has been fixed. |
|
VER-66988 |
Scrutinize |
When running scrutinize, the database password would be written to scrutinize_collection.log. This has been fixed: the "__db_password__" entry has been removed from the log file. |
|
VER-62276 |
Security |
Changes to the SSLCertificate, SSLPrivateKey, and SSLCA parameters take effect for all new connections and no longer require a restart. |
|
VER-65168 |
Security |
Before release 9.1, the default Linux file system permissions on scripts generated by the Database Designer were 666 (rw-rw-rw-). Beginning with release 9.1, default permissions changed to 600 (rw-------). With this release, default permissions have reverted to 666 (rw-rw-rw). |
|
VER-65890 |
Security |
Two system tables have been added to monitor inherited privileges on tables and views: inheriting_objects shows which catalog objects have inheritance enabled, and inherited_privileges shows what privileges users and roles inherit on those objects. |
|
VER-68616 |
Security |
When the only privileges on a view came via ownership and schema-inherited privileges instead of GRANT statements, queries on the view by non-owners bypassed the privilege check for the view owner on the anchor relation(s). Queries by non-owners with inherited privileges on a view now correctly ensure that the view owner has SELECT WITH GRANT OPTION on the anchor relation(s). |
|
VER-67658 |
Spread |
In unstable networks, some UDP-based Vertica control messages could be lost. This could result in hanging sessions that could not be cancelled. This issue has been fixed. |
|
VER-47639 |
Supported Platforms |
Vertica now supports the XFS file storage system. |
|
VER-67234 |
Tuple Mover |
The algorithm for prioritizing mergeout requests sometimes overlooked slow-loading jobs, especially when these competed with faster jobs that loaded directly into ROS. This could cause mergeout requests to queue indefinitely, leading to ROS pushback. This problem was resolved by changing the algorithm for prioritizing mergeout requests. |
|
VER-67275 |
Tuple Mover |
Previously, the mergeout thread dedicated to processing active partition jobs ignored eligible jobs in the mergeout queue, if a non-eligible job was at the top of the queue. This issue has been resolved: now the thread scans the entire queue for eligible mergeout jobs. |
|
VER-68383 |
Tuple Mover |
Mergeout did not execute purge requests on storage containers for partitioned tables if the requests had invalid partition keys. At the same time, the Tuple Mover generated and queued purge requests without validating their partition keys. As a result, mergeout was liable to generate repeated purge requests that it could not execute, which led to rapid growth of the Vertica log. This issue has been resolved. |
|
VER-65744 |
UDX |
Flex table views now properly show UTF-8 encoded multi-byte characters. |
|
VER-62048 |
UI - Management Console |
Certain design steps in the Events history tab of the MC Design page were appearing in the wrong order and with incorrect time stamps. This issue has been fixed. |
|
VER-65561 |
UI - Management Console |
Under some conditions in MC, a password could appear in the scrutinize collection log. This issue has been fixed. |
|
VER-67903 |
UI - Management Console |
MC was not correctly handling design tables when they contained Japanese characters. This issue has been resolved. |
|
VER-68357 |
UI - Management Console |
When Clockskew was detected, MC was incorrectly updating the alert during the subsequent checks with the timestamp of the last check, instead of the timestamp when the Clockskew was first detected. This issue was fixed, and also the status of the alert (resolved/unresolved) was added to the display. |
|
VER-68448 |
UI - Management Console |
MC catches Vertica’s SNMP traps and generates an alert for each one. MC was not generating an alert for the Vertica SNMP trap: CRC Mismatch. This issue has been fixed. |
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.
|
Issue |
Component |
Description |
| VER-45474 | Optimizer | When a node down, DELETE and UPDATE query performance can slow due to non-optimized query plans. |
| VER-68463 | Cloud - Amazon, Data Export, Hadoop | Export to parquet with partitions fails if any of the exported columns in the outer-most select statement is specified using "schema.table.column" notation. The workaround is to specify it using just "column" notation without "schema.table" part. |
| VER-67228 | AMI, License | An Eon Mode database created in one hourly listing can only be revived into another hourly listing with the same operating system. For example, an Eon Mode database created in the Red Hat hour listing cannot be shutdown then revived into a cluster created using the Amazon Linux listing. |
| VER-64997 | Backup/DR, Security | A vbr restore operation can fail if your database is configured for SSL and has a server.key and a server.crt file present in the catalog directory. |
| VER-64916 | Kafka Integration | When Vertica exports data collector information to Kafka via notifier, the sterilization logic does not properly handle Vertica internal data type translation. When importing or copying that data back into the Management Console, Vertica uses the raw data in Kafka and inserts garbled data into the extended monitoring database. |
| VER-64352 | SDK | Under special circumstances, a sequence of statements to create and drop Python libraries fails. For example, the following session statements attempt to create Python libraries lib1 and lib2:
CREATE LIBRARY lib1 AS :lib1_path DEPENDS :numpy_path LANGUAGE 'Python'; DROP LIBRARY lib1; CREATE LIBRARY lib2 AS :lib2_path DEPENDS :sklearn_path LANGUAGE 'Python'; Here, lib1 is a Python library that imports module numpy, while lib2 imports module sklearn, which relies on numpy. After lib1 is dropped, the CREATE LIBRARY statement for lib2 fails. |
| VER-63720 | Recovery | When fewer nodes than the node number of the cluster are specified in a vbr configuration file, the catalog of the restored library will not be installed on nodes that were not specified in the vbr configuration file. |
| VER-62983 | Hadoop | When hcatalogconnector schemas are created with custom_partitions enabled, poor performance has been observed when there are many (500+) partitions. By default, custom_partitions are disabled on hcatalogconnector schemas. |
| VER-62061 | Catalog Sync and Revive | If you revive a database on a one-node cluster, Vertica does not warn users against reviving the same database elsewhere. This can result in two instances of the same database running concurrently on different hosts, which can corrupt the database catalog. |
| VER-61584 | Nimbus, Subscriptions | It happens only while node(s) is shutting down or in unsafe status. No workaround. |
| VER-61420 | Data Removal - Delete, Purge, Partitioning | Partition operations, such as move_partitions_to_table, must split storage containers that have both partitions that match the operations and partitions that do not. Version 9.1 introduced an inefficiency whereby such a split may separate a storage container into one more storage container than necessary. Note that any extraneous containers created this way will eventually be merged by the TM. |
| VER-61069 | Execution Engine | In very rare circumstances, if a Vertica process crashes during shutdown, the remaining processes might hang indefinitely. |
| VER-60409 | AP-Advanced, Optimizer | APPLY_SVD and APPLY_PCA functions are implemented as User-Defined Transform Functions (UDTFs). A query containing a UDTF that takes many input columns (e.g. more than 1000) and also outputs many columns may fail with error "Request size too big" due to additional memory requirement in parsing. |
| VER-58168 | Recovery | A transaction that started before a node began recovery is referred to as a dirty transaction in the context of recovery. A recovering node must wait for or cancel such a transaction in order to recover tables modified by the transaction. In some rare instances such transactions may be hung and can not be canceled. Tables locked by such a transaction can not be recovered and therefore a recovering node can not transition to 'UP'. Usually the hung transaction can be stopped by restarting the node on which the transaction was initiated, assuming this is not a critical node. In extreme cases or instances where the initiator node is a critical node, restart the cluster. |
| VER-57126 | Data Removal - Delete, Purge, Partitioning | Partition operations that use a range, for example, copy_partitions_to_table, must split storage containers that span the range in order to complete. For tables partitioned using a GROUP BY expression such split plans can require a relatively large amount of memory, especially when the table has a large number of columns. In some cases the partition operation may error with "ERROR 3587: Insufficient resources to execute plan on pool <poolname>". |
| VER-48020 | Hadoop | Canceling a query that involves loading data from ORC or Parquet files can be slow if there are too many files involved in the load. |
The only warranties for products and services are set forth in the express warranty statements accompanying such products and services. Nothing herein should be construed as constituting an additional warranty. Micro Focus shall not be liable for technical or editorial errors or omissions contained herein.
The information contained herein is subject to change without notice.
Confidential computer software. Valid license from Micro Focus required for possession, use or copying. Consistent with FAR 12.211 and 12.212, Commercial Computer Software, Computer Software Documentation, and Technical Data for Commercial Items are licensed to the U.S. Government under vendor's standard commercial license.
© Copyright 2006 - 2019 Micro Focus, Inc.
Adobe® is a trademark of Adobe Systems Incorporated.
Microsoft® and Windows® are U.S. registered trademarks of Microsoft Corporation.
UNIX® is a registered trademark of The Open Group.