Greenplum Database 4.3.x Resolved Issues

A newer version of this documentation is available. Use the version menu above to view the most up-to-date release of the Greenplum 4.x documentation.

Greenplum Database 4.3.x Resolved Issues

Consolidated list of resolved issues for the Greenplum Database 4.3.x releases.

4.3.32.x Release

Table 1. Resolved Issues in 4.3.32.x
Issue Number Category Resolved In Description
29806 Server: Execution 4.3.32.1 Resolved a problem where using COPY FROM on a partitioned table caused segment failures if a partition was exchanged with a table that had dropped columns.
29560 Storage: Vacuum 4.3.32.1 Fixed a segmentation fault that could occur when autovacuuming the template0 database with the configuration parameter track_counts enabled.
29702 Storage: Transaction Management 4.3.32.1 Fixed a problem where an INSERT statement executed during a VACUUM operation could fail with ERROR: cannot insert into segno.
29772 Query Optimizer 4.3.32.1 In some cases, GPORCA crashed while incorrectly translating a correlated nested loop join (NLJ) with an empty filter expression on the outer side. This problem has been resolved. GPORCA has improved its translation of this type of correlated NLJ, and no longer tries to access an empty expression.
29815 Query Optimizer 4.3.32.1 Resolved a problem where GPORCA could not generate a plan (and therefore would not execute) a SELECT statement in a CREATE VIEW command.
29847 Query Optimizer 4.3.32.1 In some cases, GPORCA did not eliminate partitions properly due to incorrectly inferred predicates. This caused a query to run much slower with GPORCA than when run with the legacy planner. This problem has been resolved. GPORCA now pushes predicates down as far as possible before inferring.
29853 gpbackup/gprestore 4.3.32.1 When a role was assigned the permission CREATEEXTTABLE(protocol='gpfdist', type='readable'), the gpbackup utility produced a CREATE ROLE command with an incorrect protocol and a missing type argument. This issue has been resolved.
29850 Query Optimizer 4.3.32.1 In a query with a full join and a filter, GPORCA did not push down the filter to eliminate rows before performing the joins. As a result, the entire relation was scanned and used to perform the full join before the filter was applied. This processed a larger number of rows than necessary, causing the query to run much slower than when run with the legacy planner. This issue has been resolved. Queries with full outer joins now fall back to the legacy planner.
163807792 gpbackup/ gprestore 4.3.32.0 When the % sign was specified as the delimeter in an external table text format, gpbackup escaped the % sign incorrectly in the CREATE EXTERNAL TABLE command. This has been resolved. The % sign is correctly escaped.
163807792 gpbackup/ gprestore 4.3.32.0 When a user-defined function set a configuration parameter with a string value such as '1MB' or '64kB', gpbackup omitted the quotes when saving the CREATE FUNCTION command to the backup. This issue has been resolved. gpbackup now encloses the configuration value in quotes.
29679 Storage: Catalog and Metadata 4.3.32.0 Resolved an off-by-one error in transactional snapshots. In previous releases, Greenplum incorrectly pointed the first snapshot's in-progress transaction array at the second block that was allocated for transactions in progress. This caused every snapshot N to point to snapshot N+1's block of allocated transactions, and for the last snapshot to point to an address that was not allocated for transactions in progress. This has been fixed.
29716 Resource Management 4.3.32.0 When gp_vmem_protect_limit was set to 18384 or higher and the Greenplum cluster was reset multiple times, users could fail to connect to the cluster with the error "FATAL","Too many sessions.". This problem was caused by a memory management error. This was fixed by properly initializing the cluster memory setting.
29804 analyzedb 4.3.32.0 When analyzing a child partition of a partitioned table with the analyzedb utility, the root partition was not also analyzed. This has been resolved. The utility now analyzes both the child partition and the root partition.

4.3.31.0 Release

Table 2. Resolved Issues in 4.3.31.0
Issue Number Category Resolved In Description
29696 Query Planner 4.3.31.0 For some queries where the predicate contains a subquery that contains a comparison such as BETWEEN '2' AND '1', the Greenplum Database legacy optimizer incorrectly returned an error stating that no parameter was found for an initplan subquery. The legacy optimizer did not correctly handle the comparison when generating a query plan.

This issue has been resolved. The legacy optimizer has improved the handling of comparisons in subqueries.

29692 gpbackup/ gprestore 4.3.31.0 In some cases, a gprestore command that specified the --restore-globals command failed with an invalid syntax message for the timestamp infinity-00. The error occurred because gpbackup did not correctly backup a role when a role's expiration date and time is set to infinity.

This issue has been resolved. Now gpbackup correctly backs up a role's expiration date and time in the specified situation.

29487 gpdbrestore 4.3.31.0 In some cases when restoring a partitioned table, the gpdbrestore utility logged a message stating that a constraint could not be dropped during the restore operation. The message was generated by an incorrect DROP command that the utility used to drop a partitioned table before recreating it.

This issue has been resolved. The utility has been improved to not generate the message in the specified situation.

29358 Query Planner 4.3.31.0 The legacy query optimizer returned the error no relation entry for relid 0 for queries against a partitioned table, when the query contained a subquery that did not reference the partitioned table. The error occurred during query planning when the legacy optimizer did not correctly check the hierarchical relationship of a leaf table of the partitioned table.

This issue has been resolved. The checking of leaf table relationships by the legacy optimizer has been improved for the specified type of queries.

162025531 gpbackup/ gprestore 4.3.31.0 In some cases when gprestore was run with the option --include-table to restore tables that were defined with sequences, tables were not restored. The table was not restored if a sequence was altered so that the sequence was owned by the table (ALTER SEQUENCE...OWNED BY). The gpbackup utility did not back up the specified type of table and sequence correctly.

This issue has been resolved. The gpbackup utility has been improved to back up the specified table and sequence correctly.

161393615 gpbackup/ gprestore 4.3.31.0 The gpbackup utility did not correctly back up comments on an index that was created as a UNIQUE or PRIMARY KEY index.

This issue has been resolved. Now the utility correctly backs up comments on the specified type of index.

4.3.30.x Release

Table 3. Resolved Issues in 4.3.30.x
Issue Number Category Resolved In Description
29625 Catalog and Metadata 4.3.30.4 The Greenplum Database timezone files were not up to date. For example, on 2017-12-18 Brazil moved the DST date from October 21 to November 04. This was not reflected in the Greenplum Database timezone files.

This issue has been resolved. The timezone files have been updated. See Changed Features.

29616 COPY 4.3.30.4 In some cases, copying CSV data with the COPY...FROM STDIN command caused a Greenplum Database PANIC. The PANIC occurred when Greenplum Database did not correctly handle an error caused by invalid input data.

This issue has been resolved. Now Greenplum Database handles the error correctly and does not generate a PANIC.

29638 VACUUM 4.3.30.3 In some cases when a VACUUM operation was performed on an append-optimized table and other concurrent operations were also being performed on the table, Greenplum Database returned an error stating that the state for a segno from a AO relid was incorrect. The error was cause when Greenplum Database did not correctly track the segment files that had been compacted.

This issue has been resolved. Greenplum Database has improved the tracking of append-optimized segment files during a VACUUM operation.

29635 gpbackup/ gprestore 4.3.30.3 In some cases running a gprestore command with the --on-error-continue option returns an error when creating database metadata. The error is returned when the utility did not correctly handle an error returned by a CREATE SCHEMA command.

This issue has been resolved. Now running the utility with the --on-error-continue option correctly handles errors returned by CREATE SCHEMA commands.

29634 Query Optimizer 4.3.30.3 In some cases, GPORCA generated a Greenplum Database PANIC when executing queries that contain the nullif() function. The PANIC occurred when GPORCA did not return the nullif() result cast as the correct data type.

This issue has been resolved. Now GPORCA returns the function result correctly.

29622 gpcrondump/ gpdbrestore 4.3.30.3 In some cases when the gpdbrestore utility performed a restore operation using Data Domain Boost (DD Boost), the operation failed. The utility did not process backup information for creating database objects when restoring using DD Boost.

This issue has been resolved. Now the utility processes backup information for creating database objects and does not fail.

29611 Query Optimizer 4.3.30.3 For some queries with complex join conditions, GPORCA caused a Greenplum Database PANIC. The PANIC was caused when GPORCA did not correctly determine when possible joins would be cross joins when evaluating join ordering alternatives during query optimization.

This issue has been resolved. GPORCA has improved join order evaluation during query optimization.

29601 gprecoverseg 4.3.30.3 In some cases, the gprecoverseg utility generated a Greenplum Database PANIC during a segment instance recovery operation. The PANIC occurred when the utility generated a SIGSEGV when it did not manage shared memory correctly while a segment instance was being stopped.

This issue has been resolved. The utility has improved the shared memory error handling in the specified situation.

29567 Transaction Management 4.3.30.3 In some cases, a command such as CREATE TABLE did not complete and consumed a large amount of CPU resources. The issue occurred when Greenplum Database did not manage object ID (OID) values correctly when OID value wraparound occurred.

This issue has been resolved. The handling of OID values has been improved for the specified situation.

29428 Query Planner 4.3.30.3 The Greenplum Database legacy optimizer generated a PANIC for some queries that contain a window function when the window function contains a parameter. The query plan generated by the legacy optimizer did not correctly manage parameters when generating subplans for the query.

This issue has been resolved. The legacy optimizer has improved parameter management for the specified type of query.

29360 Query Execution 4.3.30.3 When the server configuration parameter gp_dynamic_partition_pruning was set to off, a query could generate a Greenplum Database PANIC when the query contained an IN clause and the clause contained a subselect. The PANIC was caused when Greenplum Database did not manage the generated IN clause values correctly.

This issue has been resolved. Greenplum Database has improved the handling of IN clause values in the specified situation.

161653513 gpdbrestore 4.3.30.3 The gpdbrestore utility returned an error when the backup that was being restored contained a comment defined on an external table. The error occurred when the utility did not correctly restore the comment.

This issue has been resolved. Now the utility restores comments defined on an external tables.

161635233 gpbackup/ gprestore 4.3.30.3 The gprestore utility returned an error when the restore command specified both the --create-db and --restore-globals options. The failure occurred when the utility attempted to restore roles before creating the database.

This issue has been resolved. Now the utility restores roles after creating the database in the specified situation.

161607786

gpbackup/ gprestore 4.3.30.3 In some cases, when a gprestore command specified the --restore-globals to restore Greenplum Database system objects such as roles, and grant assignments on roles, the utility returned an error stating that a role did not exist. The failure occurred when the utility attempted to execute a GRANT command specifying a role that did not exist.

This issue has been resolved. Now the utility restores roles before executing GRANT commands that specify a role.

161271960 gpbackup/ gprestore 4.3.30.3 In some cases, a restore operation failed when using the gprestore utility with the --include-schema, --exclude-schema, or --exclude-table options. The restore error incorrectly listed a table that was not being restored but already existed in the database. When checking for the existence of tables in the target database, the utility incorrectly checked for tables that were in the backup set but were not being restored.

This issue has been resolved. The gprestore utility has been improved to perform the restore operation without displaying an error in the specified situation.

161074289 gpbackup/ gprestore 4.3.30.3 During a backup operation, the gpbackup utility failed if a user was connected to the database that was being backed up and ran an ALTER ROLE...SET <config_parameter> command to change the default for a server configuration parameter for a role's session.

This issue has been resolved. Now the utility ignores the command during a backup operation and does not fail.

160975618 gpbackup/ gprestore 4.3.30.3 When performing a backup operation, the gpbackup utility failed if the command specified the option --exclude-table or --exclude-schema and the table or schema to be excluded did not exist.

This issue has been resolved. For the specified situation, now the utility does not fail and logs a warning that the table or schema does not exist.

160627949 VACUUM 4.3.30.3 The Greenplum Database autovacuum process incorrectly tested a database that allows connections to manage transaction ID (XID) values to avoid XID wraparound issues. The process should only test databases that do not allow connections (such as the template0 database). In some cases, testing a database that allows connections caused the process to go into an infinite loop.

This issue has been resolved. The autovacuum tests that check databases has been improved to only test databases that do not allow connections.

29594

160864241

gpbackup/ gprestore 4.3.30.0 The gprestore utility failed if the restore operation restored to a database and the name contained upper and lower case characters or special characters.

This issue has been resolved. Now the utility can perform a restore operation to a database with a valid name that contains upper and lower case characters and special characters.

29554 analyzedb 4.3.30.0 The analyzedb utility failed when the schema name or table name for the table being analyzed contains upper and lower case characters or special characters.

This issue has been resolved. The utility has been enhanced to handle the specified type of schema and table names.

29550 DML 4.3.30.0 In some cases, Greenplum Database did not properly insert data into an append-optimized table when a VACUUM operation ran concurrently with the INSERT operation. Greenplum Database did not correctly manage table files on disk that were being vacuumed.

This issue has been resolved. Greenplum Database has improved how append-optimized table disk files are managed in the specified situation.

29545 Query Optimizer 4.3.30.0 For some queries that contain a JOIN and a join column is not defined with an index, GPORCA generated a suboptimal plan that contains an Indexed Nested Loop Join. The unnecessary Index Scan performed by the Indexed Nested Loop Join caused performance issues.

This issue has been resolved. GPORCA has improved predicate checking to ensure Indexed Nested Loop Joins are related to the join conditions.

29544 gpbackup/ gprestore 4.3.30.0 When performing some backup operations, the gpbackup utility failed and returned the error Dependency resolution failed. The error occurred when the utility did not handle dependencies for some database objects correctly.

This issue has been resolved. The utility has improved the handling dependent database objects.

29530 gpdbrestore 4.3.30.0 When restoring database objects with the gpdbrestore utility with an option to restore specific objects, such the -S option to restore a specific schema, the utility did not correctly restore table columns when the column is defined as a serial data type.

This issue has been resolved. Now the utility correctly restores a column that is defined as a serial data type in the specified situation.

29526 Query Planner 4.3.30.0 For some queries that contain a nested JOIN and the inner JOIN uses a SELECT statement, Greenplum Database returned an error. The error was returned when Greenplum Database did not correctly manage the processing of the inner and outer joins.

This issue has been resolved. Greenplum Database has improved the management of joins for the specified type of queries.

29523 gptoolkit 4.3.30.0 The Greenplum Database view gp_toolkit.gp_bloat_expected_pages view might incorrectly report that a root partition table was bloated even though root partition tables did not contain data. This information could cause a user to run a VACUUM FULL operation on the partitioned table when the operation was not required.

The issue was resolved. Now the view excludes root partition tables when reporting possible table bloat. See Update for gp_toolkit.gp_bloat_expected_pages Issue.

29522 Query Optimizer 4.3.30.0 For some queries that contain a LEFT JOIN and UNION ALL, GPORCA incorrectly generated constraints while processing the LEFT JOIN. This caused the query to return incorrect results.

This issue has been resolved. GPORCA has been improved the join processing for the specified type of queries.

29521 Catalog and Metadata 4.3.30.0 In some rare cases, system catalog issues occurred after recovering from a Greenplum Database host shut down, by performing a gprecoverseg followed by a gprecoverseg -r to rebalance primary and mirror segments. The issues occurred during the rebalance operation because gprecoverseg did not properly handle change tracking entries for hint bits.

This issue has been resolved. The utility has been enhanced to properly handle the specified type of change tracking entries.

29519 Resource Queue 4.3.30.0 In some cases, when query execution was controlled by a resource queue and the query triggered a rule that runs an SQL command, Greenplum Database returned the error out of shared memory adding portal increments. The error was returned when the resource queue did not properly handle the SQL executed by the rule.

This issue has been resolved. Management of SQL commands by resource queues has been improved to handle the specified situation.

29504 gpbackup/ gprestore 4.3.30.0 In some cases, a restore operation performed with gprestore failed with the error data line too long.

This issue has been resolved. The maximum length of CSV data supported by Greenplum Database has been increased to 1GB.

29499 Query Optimizer 4.3.30.0 For queries that generate query plans that contain predicates that are inferred from the query predicates, GPORCA might generate suboptimal plans because of the placement of the inferred predicates in the query plan. This caused performance issues for some queries.

This issue has been resolved. The GPORCA algorithm that determines inferred predicates and places them in the query plan has been improved.

29493 Query Execution 4.3.30.0 In some cases when executing a query, Greenplum Database returned an invalid memory alloc request error. The error was generated by Greenplum Database while monitoring the memory used during query execution when the query executed a large number of functions. The number of functions caused a the Greenplum Database to consume a large amount of memory during query execution.

This issue has been resolved. Greenplum Database has enhanced how it monitors query execution memory to consume less memory.

29441 gpperfmon 4.3.30.0 In the query_history table of the gpperfmon database, a date in 1969 was incorrectly recorded in the tsubmit column for ALTER TABLE...SET DISTRIBUTED commands. The command is not a query and was incorrectly recorded in the table.

This issue has been resolved. The commands are no longer recorded in the query_history table.

26829 DDL 4.3.30.0 When creating a partitioned table with the CREATE TABLE command, Greenplum Database generated a PANIC when a SUBPARTITION TEMPLATE clause was not nested in a SUBPARTITION BY clause.

This issue has been resolved. Greenplum Database syntax checking for CREATE TABLE has been enhanced to return an error in the specified situation.

26589 DDL 4.3.30.0 In a database, a DROP SCHEMA operation (without the CASCADE keyword) could succeed while database objects were being created in the schema. This might have created entries in the pg_class and pg_type tables with invalid namespace values.

This issue has been resolved. Locking on schemas has been improved when creating database objects in the schema and when performing a DROP SCHEMA operation on the schema.

160999837 gpbackup/ gprestore 4.3.30.0 The gpbackup utility incorrectly backed up the table definition of external web tables when performing a backup operation. This caused gprestore to fail when attempting to restore the table definition.

This issue has been resolved. Now gpbackup correctly backs up the table definition of up external web tables.

160941596 gpbackup/ gprestore 4.3.30.0 When the gprestore utility performed a restore operation, datatypes that are defined with the storage attribute set to external were incorrectly restored with the storage attribute set to extended.

This issue has been resolved. Now the utility correctly restores datatypes that are defined with the storage attribute set to external.

159838009, 159966889 gpbackup/ gprestore 4.3.30.0 If you performed a restore operation with the gprestore utility and specified a table filtering option such as --include-table to restore a partitioned table, the partitioned table was restored, however the data was not restored.

This issue has been resolved. Now the utility restores the partitioned table and the table data in the specified situation.

159810802 gpbackup/ gprestore 4.3.30.0 If the gpbackup utility encountered a FATAL error during a backup operation, gpbackup_helper processes might not be terminated might continue running on the segment hosts.

This issue has been resolved. Now the processes are terminated on the segment hosts in the specified situation.

159605003 gpbackup/ gprestore 4.3.30.0 If the gpbackup utility encountered a FATAL error during a backup operation, the utility incorrectly displayed the warning [WARNING]: Failed to remove lock file.

This issue has been resolved. The utility does not display the warning in the specified situation.

4.3.29.0 Release

Table 4. Resolved Issues in 4.3.29.0
Issue Number Category Resolved In Description
29511 gpbackup/ gprestore 4.3.29.0 In some cases when gpbackup was performing a backup operation with the --single-data-file and --backup-dir options, the operation failed with the error Pipe not found. The backup failed when I/O performance of the backup directory file system was slow. The failure occurred when gpbackup did not properly manage copying small amounts of data to a backup directory with slow I/O.

This issue has been resolved. The utility has improved how data is copied in the specified situation.

29506 gpbackup/ gprestore 4.3.29.0 When creating a backup, the gpbackup utility did not properly quote the ROLE names in ALTER TABLE ... OWNER TO statements for tables, views and sequences. This caused a failure when gprestore attempted to restore the backup.

This issue is resolved. Now gpbackup quotes ROLE names containing upper case characters and special characters.

29486 DDL 4.3.29.0 In some cases, creating a partitioned table generates a Greenplum Database PANIC. The error occurs when a partition column is defined as a different binary compatible type than the partition value specified in the PARTITION BY clause. In this example, the column col1 is a partition column that is defined as character varying, and the partition value is cast as a text.
CREATE TABLE mypart ... ( col1 character varying, ... ) ...
   PARTITION BY RANGE(col1) ... START ('100'::text)

This issue has been resolved. Now the specified type of partitioned table is created.

29476 gpbackup/ gprestore 4.3.29.0 When running the gpbackup utility with the --backup-dir option, the backup operation failed if the backup directory contains a directory with a name that contains -1. For example, this directory structure caused a failure <my-backup-dir>/backup-1.

This issue has been resolved. Now the utility completes in the specified situation.

159600293 gpbackup/ gprestore 4.3.29.0 When a restore operation was performed with the gprestore utility and specified the --include-table and --data-only options to restore data for a table, the utility incorrectly returned an error stating that the table already exists.

This issue has been resolved. Now in the specified situation, the utility does not return an error if the table exists. The utility returns an error if the table does not exist.

4.3.28.0 Release

Table 5. Resolved Issues in 4.3.28.0
Issue Number Category Resolved In Description
29464 gpbackup/ gprestore 4.3.28.0 The gprestore utility failed restoring GRANT information when the ROLE names associated with the information contained upper and lower case characters or special characters. The failure occurred because the gpbackup utility did not correctly handle ROLE names that contained the specified types of characters during a backup operation.

This issue has been resolved. Now gpbackup correctly handles the specified type of ROLE names.

29437 Query Optimizer 4.3.28.0 For some queries against temporary tables or tables without statistics, the query plan generated by the legacy query optimizer was much more efficient than the plan generated by GPORCA. For a table without statistics, you can set the server configuration parameter gp_enable_relsize_collection to on so that the legacy optimizer uses the estimated size of a table if there are no statistics for the table. GPORCA ignored this parameter and used a default value to estimate the number of rows.

This issue has been resolved. Now GPORCA supports the parameter. See Changed Features.

29436 gpbackup/ gprestore 4.3.28.0 When backing up a view, the gpbackup utility did not include the owner of the view.

This issue has been resolved. Now the utility includes the owner information when backing up a view.

29435 gpbackup/ gprestore 4.3.28.0 The gpbackup utility failed when a backup operation was performed on a Greenplum Database system was initialized with --locale=C, the database was configured with encoding=WIN874, and the name of a table that was being backed contained a Thai character.

This issue has been resolved. Now the backup operation completes in the specified situation.

29418 External Tables 4.3.28.0 When inserting data from an external table, the INSERT command did not recognize the Greenplum Database server configuration parameter gp_max_csv_line_length. This caused INSERT commands to fail with a data line too long error when the row from the external table was shorter than the value specified by the configuration parameter gp_max_csv_line_length.

This issue has been resolved. Now the INSERT command recognizes the parameter and successfully inserts data from an external table in the specified situation.

29418 gpperfmon 4.3.28.0 When copying data from an external CSV format table, if a text column contains line breaks and one of the lines is longer than the value of the gp_max_csv_line_length server configuration parameter, the copy fails with an invalid CSV data error. The error prevents gpperfmon from loading query data from the queries_tail external table into the queries_history table when query text contains line breaks and very long lines.

The gpperfmon code has been updated to prevent the error by setting the gp_max_csv_line_length configuration parameter to its maximum value (4MB) in the session. If any line in the query exceeds this length, all line breaks in the query are replaced with spaces. Although this alters the formatting of the query text for some queries, it prevents the COPY command from failing to load the queries_history table.

29410 Query Optimizer 4.3.28.0 For some queries that contain an EXCEPT clause that includes a partitioned table and that involved a bitmap index, GPORCA generated a PANIC. GPORCA did not correctly generate some alternative plans during optimization. The incorrect plan generation caused the PANIC.

This issue has been resolved. GPORCA plan generation has been improved for the specify type of queries.

29405 PL/Java 4.3.28.0 In some cases a PL/Java error generated a Greenplum Database PANIC.

This issue has been resolved. The PL/Java extension package has been updated and includes JRE 1.8.0 with update 172 that resolve this issue. See Changed Features.

29357 Storage: Access Methods 4.3.28.0 In some cases, deleting data from an append optimized tables returned an error stating that Greenplum Database could not find segment file to use. The error was caused when the cached state of an append-optimized table was not handled correctly during an ALTER TABLE...SET DISTRIBUTED BY or TRUNCATE operation.

This issue has been resolved. The handling of the cached state of an append-optimized table row has been improved.

29347 pgbouncer 4.3.28.0 When Greenplum Database is configured to use the PgBouncer connection pool manager, PgBouncer occasionally returned an invalid server parametererror. The error was caused when PgBouncer did not handle SET commands correctly.

This issue has been resolved. PgBouncer that ships with Greenplum Databasehas been updated to verson 1.8.1 and handles SET commands correctly. See Changed Features.

29304 Query Optimizer 4.3.28.0 For some queries against tables that are defined with a multi-key index where the query could use the index, the query plan generated by GPORCA was slower than the plan generated by the legacy query optimizer.

GPORCA has been enhanced to generate a query plan that better utilizes multi-key indexes. Now GPORCA considers plans with index alternatives that include the maximum number of predicate columns and exclude the multi-key indexes whose leading key is not present in the predicate.

159211957 gpbackup/ gprestore 4.3.28.0 In some cases, the gprestore utility restored table data that caused a Greenplum Database PANIC. The issue was caused when, in come cases, the gpbackup utility version 1.0.x backed up table data with the incorrect distribution policy and then gprestore version earlier than 1.6.1 restored the table data without performing data distribution checking to ensure data is distributed to segment instances correctly.

This issue has been resolved. gpbackup version 1.1 and later backs up table data with the correct distribution policy. Also, gprestore version 1.6.1 and later performs data distribution checking.

159125797 Documentation 4.3.28.0 In the Greenplum Database 4.3.27.0 documentation, the PgBouncer information incorrectly documented PgBouncer 1.8.1.

The Greenplum Database 4.3.28.0 documentation for PgBouncer documentation is correct. PgBouncer that ships with Greenplum Database 4.3.28.0 has been updated to 1.8.1. See Changed Features.

158574172 gpbackup/ gprestore 4.3.28.0 The gprestore utility restored external table definitions before restoring the required protocols. This caused a restore failure.

This issue has been resolved. Now the utility restores protocols before the external table definitions.

4.3.27.0 Release

Table 6. Resolved Issues in 4.3.27.x
Issue Number Category Resolved In Description
29415 gpbackup/ gprestore 4.3.27.0 In some cases, the gpbackup utility did not correctly release locks on some tables when the --include-schema option was specified and the backup included partitioned tables. In some cases, the incorrect locking caused table access issues.

This issue has been resolved. Table locking during back ups performed by gpbackup has been improved.

29374 gpinitstandby/ gpactivatestandby 4.3.27.0 In some cases, long running Greenplum Database utilities that use SSH failed to complete. To fulfill operating system hardening requirements, some systems require an SSH connection to be dropped if it has not been used for a period of time. This caused some long running Greenplum Database utilities to fail when the operating system dropped the SSH connection.

This issue has been resolved. The Greenplum Database utilities have been enhanced to keep the SSH connection active until the utility completes.

29372 gpcrondump/ gpdbrestore 4.3.27.0 When the gprestore utility attempted to restore an external table and the --truncate option was specified, the utility failed when it attempted to truncate the external table.

This issue has been resolved. Now the utility does not attempt to truncate external tables in the specified situation.

29364 Query Optimizer 4.3.27.0 GPORCA performance was poor for some queries that required a large number of joins.

GPORCA has improved the use join cardinality estimates during join order processing when creating a query plan.

29362 gpcrondump/ gpdbrestore 4.3.27.0 In some cases, backup performance was poor when performing a backup with the gpcrondump utility and specifying --table-file option. The method that the utility used to check for tables to be backed up was inefficient.

This issue has been resolved. The method used to check for tables to be backed up has been improved.

29302 Query Optimizer 4.3.27.0 GPORCA incorrectly handled the comparison operators ANY, ALL, EXISTS, or NOT-EXISTS that occured in subqueries that were nested inside another scalar expression. This resulted in an incorrect query plan that produced incorrect results.

This issue has been resolved. Now GPORCA generates a correct plan for specified type of query.

158731322 gpbackup/ gprestore 4.3.27.0 When backing up user defined data types, the gpbackup utility did not include precision information for numeric types.

This issue has been resolved. Now the backup includes the precision information.

158574185 gpbackup/ gprestore 4.3.27.0

A back up operation failed when the gpbackup utility was run with the --single-data-file option and the backup attempted to back up an external table.

This issue has been resolved. Now the back up operation completes and the backup includes the external table DDL.

158468023 gpbackup/ gprestore 4.3.27.0 When performing a back up operation, the gpbackup utility did not exclude all database objects created when a Greenplum Database extension was installed. This caused errors when restoring the backup with the gprestore utility.

This issue has been resolved. Now the gpbackup utility excludes all objects created by extensions. The objects are created when the extension is restored.

158265682 gpbackup/ gprestore 4.3.27.0 When backing up indexes, the gpbackup utility did not include the schema when backing up some index information. This caused errors when restoring the index with the gprestore utility.

This issue has been resolved. Now the gpbackup utility qualifies the index information when backing it up.

4.3.26.0 Release

Table 7. Resolved Issues in 4.3.26.x
Issue Number Category Resolved In Description
4938 Query Optimizer 4.3.26.0 GPORCA generated a SIGSEGV when processing a LIKE clause to derive additional constraints and predicates. The GPORCA constraint derivation framework did not handle the LIKE clause correctly.

This issue has been resolved. The GPORCA constraint derivation framework has been enhanced to ignore unsupported features such as the LIKE clause.

4909 pg_dump 4.3.26.0 The Greenplum Database pg_dump utility incorrectly included Greenplum Database protocol information when performing a backup operation of tables or schemas (for example, when running the command pg_dump -s -t).

This issue has been resolved. The utility no long includes protocols when backing up tables or schemas.

29392 gpbackup/ gprestore 4.3.26.0 Restoring a table with the gprestore utility, when the table depends on a sequence and the --include-table-file option was specified to restore the table.
This issue has been resolved. The gpbackup and gprestore utilities have been enhanced to support restoring dependent objects such as sequences in the specified situation when the --include-table-file or --include-table option is specified.
Note: You must explicitly specify the dependent sequence when backing up or restoring a table with the --include-table-file or --include-table option. See New Features.
29387 Query Optimizer 4.3.26.0 GPORCA generated a PANIC for some queries that contain a WITH ... UNION ALL clause and a computed column that is used in the WITH clause. GPORCA did not perform a predicate pushdown correctly during the query optimization process.

This issue has been resolved. GPORCA has been enhanced to perform predicate pushdown correctly for the specified type of query.

29385 Query Optimizer 4.3.26.0 For some queries that contain a negation comparison (comparison with a NOT operator prefix) that references multiple columns, GPORCA generated a Greenplum Database PANIC. For example, the negation comparison NOT(a.f2=0 OR a.f3=0) references two columns, f2 and f3. GPORCA did not correctly evaluate the comparison and caused a PANIC.

This issue has been resolved. GPORCA has been enhanced to correctly evaluate the comparison.

29373 gpbackup/ gprestore 4.3.26.0 The gprestore utility failed when attempting to restore Greenplum Database roles when the role name started with a digit. The gpbackup utility did not correctly handle the role during a backup operation.

This issue has been resolved. Now the utilities correctly handle the specified type of roles.

29352 gpbackup/ gprestore 4.3.26.0 Performing a gprestore operation failed if the backup contained database option values that were strings longer than 64 characters. For example, Greenplum Database server configuration parameter gp_default_storage_options can be set in a database and can have a value that is longer than 64 characters.

This issue has been resolved. The handling of long strings by gpbackup and gprestore has been improved to allow strings longer than 64 characters.

29334 gptransfer 4.3.26.0 The Greenplum Database gptransfer utility did not copy a sequence when the sequence was specified as a default value. The utility copied a sequence that is created when a table is defined with a column with a serial data type.

This issue has been resolved. Now the utility transfers the specified sequences for tables.

29323 Transaction Management 4.3.26.0 On Greenplum Database, a deadlock occurred if simultaneous INSERT and TRUNCATE operations were performed on the same table. The issue was caused by incorrect lock checking.

This issue has been resolved. Greenplum Database lock checking has been enhanced to avoid deadlocks in the specified situation.

29322 Query Execution 4.3.26.0 On Greenplum Database, running a query that executed the function gp_partition_selection generated a PANIC. The function is not supported in Greenplum Database.

Now Greenplum Database returns an error stating that the function is not supported when it is called.

29308 Catalog and Metadata 4.3.26.0 In some situations, Greenplum Database did not correctly manage adding partitions to a partitioned tables and returned a too many partitions error.

This issue has been resolved. Greenplum Database management of partitioned tables when adding partitions has been enhanced.

29307 Query Planner 4.3.26.0 For some queries against a partitioned table, Greenplum Database generated a PANIC when generating a query plan. The legacy query planner did not correctly access column statistics from a leaf table of a partitioned table if a column was deleted from the partitioned table and then a partition was split.

This issue has been resolved. The legacy query planner has been enhanced to correctly access column statistics for the specified situation.

29295 Query Optimizer 4.3.26.0 For some queries against partitioned tables that contain a UNION ALL and a subselect, GPORCA optimization took a significant amount of time when compared to the legacy query optimizer. During query planning, GPORCA did not correctly manage scalar subqueries when generating a query plan and caused a long query plan generation time.

This issue has been resolved. In GPORCA management of scalar subqueries has been enhanced for the specified type of queries.

29288

29312

gptransfer 4.3.26.0 The gptransfer utility did not correctly copy table sequences (sequences that are created when a table column is defined with the data type SERIAL). The utility reset the sequence last_value field to 1.

This issue has been resolved. Now the utility correctly copies table sequences.

29284 PL/Python 4.3.26.0 Queries running on a Greenplum system that were executing a PL/Python function could not be cancelled using the functions pg_cancel_backend or pg_terminate_backend. The PL/Python function did not recognize a cancel signal from pg_cancel_backend.

This issue has been resolved. Greenplum Database PL/Python has been enhanced to recognize a cancel signal from pg_cancel_backend.

158093886 gpbackup/ gprestore 4.3.26.0 During a backup operation, the gpbackup utility did not properly qualify external tables with a schema when the table was defined with an error table. This caused the gpbackup utility to restore the table in the incorrect location.

This issue has been resolved. Now the utility properly qualifies the external table.

157550115 COPY 4.3.26.0 A COPY ... ON SEGMENT command reported an incorrect number of rows when the command copies more than 231 rows. This issue occurred because row count used a 32-bit integer. This issue affected Greenplum Database utilities that used COPY ... ON SEGMENT, such as gpcopy.

This issue has been resolved. The row count has been enhanced to use a 64-bit integer.

156493482 pg_dump 4.3.26.0 When using the Greenpum Database pg_dump utility, the backup operation failed if the database contains tables with names that contain special characters, such as a single quote ('). For example, the name test'tbl contains a single quote.

This issue has been resolved. Now the utility can backup tables with names that contain special characters.

156027741 Query Optimizer 4.3.26.0 GPORCA generated a PANIC when accessing statistics for columns and the statistics data types do not match the column type defined by the DDL.

Now in the specified situation, GPORCA creates a default set of statistics and issues a warning to perform an ANALYZE operation to correct the issue.

4.3.25.x Release

Table 8. Resolved Issues in 4.3.25.x
Issue Number Category Resolved In Description
29330 gphdfs 4.3.25.1 In some cases, an external table defined with the gphdfs protocol generated an error when reading a Parquet file. The error occurred when handling complex types with optional fields and the Parquet data did not contain data for the optional field.

This issue has been resolved. The gphdfs protocol has been enhanced to handle the specified type of Parquet data.

29299 gpcrondump/ gpdbrestore 4.3.25.1 The Greenplum Database gpdbrestore utility failed when restoring a backup from a Data Domain Boost system to a system where the Greenplum Database DBID information has changed. For example, a segment instance failed over to a mirror segment between the backup and restore operation.

This issue has been resolved. The gpdbrestore utility has been enhanced to restore backups from Data Domain Boost system in the specified situation.

29283 Query Optimizer 4.3.25.1 GPROCA optimization performance has been improved for queries that contain a predicate that performs a comparison with an constant array with a large number of elements where the array elements are sets of constants.

This issue has been resolved. GPROCA optimization performance has been enhanced for the specified type of queries. See Changed Features.

29275 Query Optimizer 4.3.25.1 For some multiple join queries, GPORCA incorrectly disabled the minimum cardinality based join ordering whenever the join associativity threshold was met to improve optimization processing. This caused GPORCA to generate a suboptimal plan for some multiple join queries.

This issue has been resolved. GPORCA has improved join processing for multiple join queries.

29255

27037

Query Planner, Query Optimizer 4.3.25.1 In some cases, both the Greenplum legacy planner and GPORCA generated a SIGSEGV during query plan generation. The SIGSEGV occurred when memory for the statistics used in the query was not managed correctly when there are issues with the statistics.

This issue has been resolved. Now memory management for statistics has been improved and a plan is generated.

26529 Query Optimizer 4.3.25.1 For GORCA, cardinality for a join that includes a predicate of the form f(a) = b was estimated using a constant scale factor. This estimate resulted in an overestimation of the cardinality, and in some cases, choosing an inferior query plan.

This issue has been resolved. GPORCA has been improved to generate cardinality estimate of the specified type of predicate and results in a better plan being selected.

29332

29333

Query Planner 4.3.25.0 In some cases, with queries that contained a common table expression (CTE) and the CTE contained a filter, the Greenplum Database legacy query planner generated a Greenplum Database PANIC during query optimization. When the server configuration parameter gp_inline_simple_cte is off or if a CTE could not be inlined, the PANIC was generated when the legacy planner did not perform a predicate pushdown correctly.

This issue has been resolved. Performing predicate pushdowns by the legacy planner has been improved for the specified type of queries.

29316 gpbackup/ gprestore 4.3.25.0 The gprestore utility failed when attempting to restore a database when the PL/Java language was installed in the database. The gpbackup utility did not correctly back up the installed PL/Java language.

This issue has been resolved. Now gpbackup correctly backs up the installed PL/Java language.

29301 gpbackup/ gprestore 4.3.25.0 In some cases, when using the gprestore utility to restore tables that owned sequences, the utility returned an error. The gpbackup utility did not correctly backup sequences when backing up a set of tables (including or excluding tables) and some of the tables owned sequences.

This issue has been resolved. Now gpbackup correctly backs up sequences that are owned by tables in the specified situation.

29271 gptransfer 4.3.25.0 The gptransfer utility failed when it attempted to copy a temporary table. Temporary tables are created in a temporary system catalog schema with the prefix pg_temp.

This issue has been resolved. Now gptransfer ignores tables in catalog schemas with the prefix pg_temp.

29259 gprecoverseg 4.3.25.0 There were gprecoverseg performance issues when the utility was run to recover a large number of Greenplum Database segment instances. Some gprecoverseg tests were performed sequentially on segments.

This issue has been resolved. To improve performance, some tests performed by the utility are now performed on segments instances in parallel.

4.3.24.0 Release

Table 9. Resolved Issues in 4.3.24.0
Issue Number Category Resolved In Description
4529 Query Optimizer 4.3.24.0 GPORCA generated an invalid plan for some queries that contain a non-redistributable column when the query contains a join on the column, or when the DISTINCT keyword is applied to the column, or when an aggregate includes the column. The invalid plan returned an error during query execution.

This issue has been resolved. GPORCA has been improved to not generate plans that contain a redistribute motion on a non-redistributable column.

29265 Query Optimizer 4.3.24.0 For some queries that qualify tables with the ONLY keyword, GPORCA returned incorrect results.

This issue has been resolved. Now GPORCA falls back to legacy planner and returns the correct results for the specified type of query.

29257 Backup/ Restore 4.3.24.0 A restore operation performed by the gprestore utility failed when the backup contained a table column with comment text and the text contained a single quote ('). The utility did not correctly handle the single quote.

This issue has been resolved. Now the utility correctly handles comment text that contains single quotes.

29249 Query Optimizer 4.3.24.0 GPORCA produced wrong results for some queries when the query contains multiple predicates combined by AND or OR operators and where some of the predicates included a NOT operator (NOT condition) . GPORCA did not process the NOT operator correctly.

This issue has been fixed. GPORCA has been enhanced to correctly process the NOT operator for the specified type of query.

29243 gpfdist/ gpload 4.3.24.0 In some cases, the gpfdist utility did not complete (hung) when accessing a named pipe and no read or write activity occurs. The gpload utility did not properly terminate a gpfdist process that has completed processing data.

This issue has been resolved. Thegpload termination process for gpfdist processes that have completed processing data has been improved. Also, the gpfdist utility now supports the GPFDIST_WATCHDOG_TIMER environment variable to help debug gpfdist hangs. See Changed Features.

29235 Query Optimizer 4.3.24.0 For some queries with multiple joins, GPORCA was incorrectly producing a distribution specification of a join that contained circular references. This caused GPORCA to go into an infinite loop and caused Greenplum Database to generate a segmentation fault.

This issue has been resolved. GPORCA has been enhanced to avoid distribution specifications with circular references.

29206 Query Planner 4.3.24.0 For some queries that contained window functions and predicates, legacy query planner created a query plan that did not push the predicate below the window function when appropriate. This caused a performance regression.

This issue has been resolved. The predicate pushdown evaluation performed by query planner has been improved for the specified type of query.

29201 Query Optimizer 4.3.24.0 For some queries against partitioned tables, GPORCA returned a row count but did not return any data. This was caused when GPORCA generated a query plan that incorrectly materialized a join that contained a dynamic table scan.

This issue has been resolved. GPORCA has been improved to avoid generating a plan that materializes a join in the specified situation.

29089 gpssh 4.3.24.0 The Greenplum Database utilities gpssh-exkeys and gpseginstall failed when they were run on a host system running RHEL 7.4 with FIPS enabled. The utilities did not support the FIPS-compatible SSH key exchange algorithm.

This issue has been resolved. The Python paramiko library in Greenplum Database Python installation has been upgraded to support newer SSH key exchange algorithms. See Changed Features.

156019082 Query Planner 4.3.24.0 The Greenplum Database legacy query planner did not process the NOT operator correctly for queries that contain an expression of the form NOT (expression) in the WHERE clause, and the expression contains SELECT statement.

This issue has been resolved. Now the legacy query planner processes the NOT operator correctly for the specified type of query.

4.3.23.0 Release

Table 10. Resolved Issues in 4.3.23.0
Issue Number Category Resolved In Description
29251 Interconnect 4.3.23.0 In some cases during query execution, a Greenplum Database segment instance generated a PANIC due to incorrect permissions on the segment host.

This issue has been resolved. Now the query fails in the specified situation when Greenplum Database encounters incorrect permissions on the segment host, the segment instance does not generate a PANIC.

29245 gpcrondump 4.3.23.0 In Greenplum Database 4.3.19.0, a performance regression was discovered in the Greenplum Database gpcondump utility. A change to a SQL command that is used by the pg_dump utility caused the performance regression.

This issue has been resolved. The SQL command has been optimized and performance has been improved.

29239 Backup and Restore 4.3.23.0 When restoring data from a backup, the gpdbrestore utility failed when Greenplum Database segment instance dbid values changed. For example, segment dbid value would change after a segment failover to a mirror.

This issue has been resolved. The restore operation succeeds in the specified situation when the segment instance contentid values are correct.

29237 Query Optimizer 4.3.23.0 GPORCA generated a Greenplum Database PANIC when processing a query where the UNICODE characters in metadata names, such as table names, are not compatible with the host system locale.

The issue has been resolved. Now GPORCA falls back to the legacy planner in the specified situations.

29226 Query Optimizer 4.3.23.0 When executing a query plan generated by GPORCA, the query failed when the query contains aggregate functions that take set operators as is input arguments. GPORCA did not recognize that this type of query was not supported.

The issue has been resolved. Now GPORCA recognizes that the specified type of query is not supported and falls back to the legacy planner.

29204 Access Methods 4.3.23.0 In some cases, a query against an Append Optimized columnar table returned an error that stated a negative bitmapset member was not allowed. This issue was caused by a data type mismatch when retrieving the first row number from the table.

The data type mismatch issue has been resolved.

29197 gpload/ gpfdist 4.3.23.0 When running a gpload operation, the gpfdist utility did not recognize a \N as the NULL character when the gpload configuration file specified a null character null_as: '\N'. When processing the configuration file, the gpload utility incorrectly escaped the backslash (\) with another backslash.

This issue has been resolved. The gpload utility has been improved to properly handle a backslash when processing the null_as property.

29194 Query Execution 4.3.23.0 In a session when the RESET or RESET ALL command was issued to reset a sever configuration parameter, the parameter was not reset for the query executer (QE) processes executing query plan slices that were running in the session.

This issue has been resolved. Now the command resets the parameter in the running query executer (QE) processes.

29193 Query Optimizer 4.3.23.0 For queries that contain a correlated subquery in the project list when legacy query optimizer generated a correlated execution plan a small percentage of such correlated subplans are on top of hash join, there is a potential of having deadlocks.

This issue has been resolved. The legacy optimizer has been improved to detect and avoid possible deadlocks for the specified type of query.

29144 COPY 4.3.23.0 The FORCE QUOTE clause of the COPY command did not enclose non-NULL data in quotes when used with a COPY TO command to copy data in CSV format.

This issue has been resolved. The data is enclosed in quotes in the specified situation.

4.3.22.0 Release

Table 11. Resolved Issues in 4.3.22.0
Issue Number Category Resolved In Description
29219 gpbackup/ gprestore 4.3.22.0 In some cases, the Greenplum Database utility gprestore would fail when restoring tables from a backup. The failure would occur when a backup was created with Greenplum Database utility gpbackup and the table exclusion option -exclude-table or -exclude-table-file was specified. The backup did not contain the schema metadata for tables included in the backup.

This issue has been resolved. The backup created in the specified situation now contains schema information.

29192 gpbackup 4.3.22.0 In some cases, the Greenplum Database gpbackup utility returned a slice bound out of range error. The error was caused when utility did not properly handle external table definitions during a backup operation.

This issue has been resolved. Now the utility handles external table definitions correctly.

29189

29043

Query Optimizer 4.3.22.0 For some queries that could use a bitmap index, GPORCA generated a sub-optimal plan that did not utilize the bitmap index.

This issue has been resolved. GPORCA has been enhanced to utilize bitmap indexes in more classes of queries.

29178

154194698

Segment Mirroring 4.3.22.0 In some situations with segment mirroring enabled, Greenplum Database did not fail over a primary segment to the mirror segment. The database would become be unusable until a manual intervention was performed. The issue was caused when the FTS (Fault Tolerance Server) process did not properly handle an error condition caused by hardware issues.

The issue has been resolved. The FTS process has been enhanced to handle the error condition.

29154 Vacuum 4.3.22.0 In some cases during a VACUUM FULL operation, Greenplum Database returned an error stating that an updated tuple was already HEAP_MOVED_OFF. The error occurred when the VACUUM operation did not correctly determine which old tuples could be deleted.

This issue has been resolved. How VACUUM checks whether old tuples can be can be deleted has been improved.

29125 Query Optimizer 4.3.22.0 In some cases, GPORCA did not properly handle a NULL value while performing contradiction detection involving text or character data types. This caused a SIGSEGV on the Greenplum Database master.

This issue has been resolved. GPORCA has been enhanced when performing contradiction detection when text or character data types have NULL values.

29062 Query Optimizer 4.3.22.0 GPORCA ignored data type modifiers that were included during query parsing and let query execution determine data type modifications. For example, the columns in this CREATE TABLE statement are defined with the data types character(10) and numeric(5,2) that modify the data types character and numeric.
CREATE TABLE mytest(col1 character(10), col2 numeric(5,2))

GPORCA discarded the data type modifications during query processing. In some cases, this caused issues when results are expected as a non-default data type.

This issue has been resolved. Now GPORCA has been improved to handle data type modifiers in table columns, scalar identifiers, constants, and length-coercion casts.

154732427 Query Optimizer 4.3.22.0 For some queries with a left outer join that contains a predicate with the IS DISTINCT FROM clause, GPORCA generated a plan that resulted in wrong results.

This issue has been resolved. GPORCA has been enhanced to return correct results for the specified type of query.

154424446 Query Optimizer 4.3.22.0 For some queries with a large number of joins, GPORCA generated a suboptimal query plan that used nested loop joins instead of hash joins. In previous versions of GPORCA, query plans that used hash joins were considered.

This issue has been resolved. Now GPORCA considers plans that use hash joins for the specified type of queries.

153757596 Query Optimizer 4.3.22.0 GPORCA did not estimate the cardinality correctly when there is a comparison between a scalar ident and an array that consists of constants and non-constants.

This issue has been resolved. Cardinality estimation has been improved for the specified situation.

4.3.21.0 Release

Table 12. Resolved Issues in 4.3.21.0
Issue Number Category Resolved In Description
29186 Query Execution 4.3.21.0 Greenplum Database generated an ERROR on a Greenplum Database segment when Greenplum Database aborted a transaction when a query executor process received invalid numeric input. The error was caused by improper memory handling during query workfile cleanup after the transaction was aborted.

This issue has been resolved. Memory handling for the specified situation has been improved.

29182 Transaction Management 4.3.21.0 In some cases, a VACUUM operation did not complete a transaction when vacuuming a database object and caused an orphan prepared transaction. The orphan prepared transaction caused issues when attempting another VACUUM operation on the same object, and after activating a primary segment after a failover. The issue was caused when the VACUUM operation did not handle transaction failures properly.

This issue has been resolved. The handling of failed transactions by VACUUM has been improved.

29167 Query Planner 4.3.21.0 For an UPDATE or DELETE command on a partitioned table that contains a subquery and a filter on a relation that appears in the subquery, the legacy query planner generated a plan with an incorrect filter for the relation inside the subquery. This caused a Greenplum Database PANIC.

This issue has been resolved. Now the legacy planner generates a valid plan.

29095 Query Optimizer 4.3.21.0 GPORCA fell back to the legacy planner when translating a query plan that contains a CAST on a subplan.

This issue has been resolved. Now GPORCA handles the specified type of query plan.

26781 PL/Java 4.3.21.0 If the Greenplum Database server configuration parameter pljava_classpath_insecure was set to true, a database user without superuser privileges could not change the server configuration parameter pljava_classpath.

This issue has been resolved. Now the parameter setting pljava_classpath_insecure=true allows non-superusers to change pljava_classpath.

153757596 Query Optimizer 4.3.21.0 GPORCA did not estimate the cardinality correctly when there is a comparison between a scalar ident and an array that consists of constants and non-constants.

This issue has been resolved. Cardinality estimation has been improved for the specified situation.

136132985 Query Optimizer 4.3.21.0

For some queries that contain correlated subqueries (CSQ), GPORCA returned incorrect results. GPORCA did not properly handle a subquery that contains both a filter such as exists or not in and an outer reference.

This issue has been resolved. The handling of the specified type of CSQ has been improved.

4.3.20.0 Release

Table 13. Resolved Issues in 4.3.20.0
Issue Number Category Resolved In Description
29187 gpfidst 4.3.20.0 In some cases, queries that contain a LIMIT clause against a readable external table that is defined with the gpfidst protocol generated a Greenplum Database segment PANIC. The error was caused when the gpfdist utility did not properly manage the end of data when the query returned the number or rows specified by LIMIT clause.

This issue has been resolved. Handling the end of data for the specified situation has been improved for the gpfdist utility.

29177 Query Optimizer 4.3.20.0 For some queries that contained a WITH clause and a projection list with more than 1 column, GPORCA returned column data in the incorrect order. The incorrect column order occurred when GPORCA did not correctly translate a ValueScan operation.

This issue has been resolved. Now GPORCA handles ValueScan operations correctly for the specified type of query.

29164 Vacuum/ Reindex/ Truncate 4.3.20.0 In some cases, Greenplum Database returned incorrect results from an append-optimized columnar table. The issue occurred after an aborted INSERT operation was followed by an ALTER TABLE ... ADD COLUMN command, and the aborted INSERT was the first INSERT operation into the table. Greenplum Database did not correctly manage the table row numbers in the specified situation.

This issue has been resolved. Now Greenplum Database manages the table row numbers correctly in the specified situation.

26518 gp_toolkit 4.3.20.0 In some cases when performing a query against the gp_toolkit.gp_bloat_diag view, Greenplum Database returned an integer out of range error. The error was caused by the incorrect handling of a column data type.

This issue has been resolved. The handling of the data type has been corrected in the gp_toolkit schema.

Note: Fixing this issue requires updating the gp_toolkit schema. See Update for gp_toolkit.gp_bloat_diag Issue.
29075 Query Optimizer 4.3.20.0 The time required by GPORCA to execute some queries against partitioned tables is much longer than for the legacy planner. The issue occurs for some queries against only partitioned tables that involve joins on the partition key of partition columns. During optimization, GPORCA generated and evaluated a large number of query plans that involved different partition selection placement options.

This issue has been resolved. The number of plans generated by GPORCA has been reduced with a minimal impact on producing an optimal query plan.

3964 Query Execution 4.3.20.0 When a COPY PROGRAM ... ON SEGMENT command was waiting for data, the command could not be terminated from a command line prompt.

This issue has been resolved. Now a COPY PROGRAM ... ON SEGMENT command can be terminated.

268 Query Optimizer 4.3.20.0 When optimizing some queries, GPORCA ignored the value of the server configuration parameter optimizer_join_order_threshold when the value was less than 10. The parameter was overridden during query optimization.

This issue has been resolved. Now GPORCA does not override the parameter.

153617158 Query Optimizer 4.3.20.0 For some queries that involve tables joins, GPORCA generated a sub-optimal plan that included a cross product on the joined tables.

This issue has been resolved. The ability of GPORCA to recognize joins and generate plans for the specified type of queries has been improved.

153541958 Query Optimizer 4.3.20.0 In some cases, GPORCA did recognize that statistics existed for a column and computed incorrect cardinality when optimizing a filter operation on a column. This caused GPORCA to generate a sub-optimal query plan.

This issue has been resolved. The recognition of column statistics by GPORCA has been improved.

153378457 Query Optimizer 4.3.20.0 In some cases, GPORCA performed query optimization when GPORCA was disabled with the server configuration parameter setting optimizer = off. The issue occurred because Greenplum Database did not correctly manage query optimization for query executors (QE) associated with the query dispatcher (QD).

This issue has been resolved. Now Greenplum Database restricts using ORCA to only QD processes running on the master segment instance.

151693327 Query Optimizer 4.3.20.0 For some queries, GPORCA did not correctly optimize a query when the DISTINCT keyword appears in a subquery that contains a [NOT] IN or [NOT] EXISTS clause. This caused GPORCA to generate a sub-optimal query plan.

This issue has been resolved. Query optimization has been improved for the specified type of query.

4.3.19.0 Release

Table 14. Resolved Issues in 4.3.19.0
Issue Number Category Resolved In Description
29143 S3 4.3.19.0 In some cases when inserting a large amount of data to an external table defined with s3 protocol, the operation failed when AWS returned an RequestTimeout error. The s3 protocol did not retry writing to the S3 data source.

This issue has been resolved. The s3 protocol has been enhanced to retry writing to the AWS S3 data store when an RequestTimeout error occurs.

29132 S3 4.3.19.0 When inserting data into a Greenplum Database writable external table that is defined with the s3 protocol and is configured to use compression, the data was not stored in compressed format in the S3 data store.

This issue has been resolved. Now the data is stored in compressed format in the S3 data store.

29129 Query Execution 4.3.19.0 In some cases, when Greenplum Database memory consumption was extremely high, the runaway query termination mechanism that is used to free memory caused a Greenplum Database PANIC. The query termination mechanism did not terminate some query sessions correctly.

This issue has been resolved. The selection and termination methods for the termination mechanism have been improved.

29008 Query Optimizer 4.3.19.0 For some queries that could use a Bitmap index, GPORCA did not choose a bitmap index scan because an incorrect cost model was used to cost bitmap scans. This caused GPORCA to choose a less efficient query plan.

This issue has been resolved. GPORCA now uses a cost model that chooses a plan with a bitmap index scan whenever applicable.

152456087 Query Optimizer 4.3.19.0 For queries against only partitioned tables where the number of table joins is greater than 2, and the exhaustive join ordering algorithm was disabled, GPORCA generated a plan that did not use Hash Join based implementation of joins.

This issue has been resolved. GPORCA now generates plans that use a more efficient Hash Joint in the specified situation.

152455099 Query Optimizer 4.3.19.0 For some queries, GPORCA generated join ordering alternatives that were swaps of join predicates rather than a change of the order of the join participants. These join ordering alternatives degraded GPORCA performance by causing unnecessary plan alternatives to be generated.

This issue has been resolved. Now GPORCA does not generate join ordering alternatives that are swaps of join predicates.

152327383 Backup and Restore 4.3.19.0 The Greenplum Database backup and restore utilities gpcrondump, gpdbrestore, pg_dump, and pg_restore could not back up and restore single-level partitioned tables partitioned tables that have been altered by exchanging a leaf child partition with a readable external table only when the partitions are named partitions (partitionname in the pg_partitions system view is not NULL for the partition).

This issue has been resolved. The utilities support backing up and restoring the specified type of table.

151530276 Query Optimizer 4.3.19.0 When generating a query plan, GPORCA did not recognize requests to interrupt and abort query optimization when processing queries that have long optimization time resulting in the VMEM limit being reached or the system becoming unstable.

This issue has been resolved. Now GPORCA aborts query optimization when it receives an interrupt request in the specified situations.

4.3.18.0 Release

Table 15. Resolved Issues in 4.3.18.0
Issue Number Category Resolved In Description
29117 Query Planner 4.3.18.0 When an INSERT command contains a subquery that contains a window function, and the INSERT predicate contains a qualifier that refers to the window function, the legacy optimizer incorrectly pushed down the qualification and generated an error.

This issue has been resolved. In the legacy optimizer, checking for predicate push down has been improved.

29110 Catalog and Metadata 4.3.18.0 In some cases, after splitting a leaf partition of a multi-level partitioned table with the ALTER TABLE command with the PARTITION SPLIT clause, inserting data into the table returned the error no partition for partitioning key. The insert error occurred because Greenplum Database did not update the partitioned table metadata correctly when splitting the leaf partition.

This issue has been resolved. Greenplum Database has improved the managing of table metadata when splitting a leaf partition.

29099 Backup and Restore 4.3.18.0 The Greenplum Database gpdbrestore utility failed when the table data contains an SQL command. For example, the text SET search_path = demo has been inserted into a table that has been backed up. The gpdbrestore utility fails when it attempted to restore the table.

This issue has been resolved. Now the utility can restore tables that contain SQL commands.

29092 Query Optimizer 4.3.18.0 GPORCA generated a PANIC when generating a plan for query that contains a function that is nested in a window function. Greenplum Database does not support queries that contain a function that is nested in a window function.

This issue has been resolved. Now Greenplum Database returns an error for the specified type of queries.

29083 DLL 4.3.18.0 If the default table storage options specified by the server configuration parameter gp_default_storage_options are different than the storage options that are defined for a table, altering the table with an ALTER TABLE command with the REORGANIZE=TRUE clause, might corrupt the table data. For example, after creating a table, changing gp_default_storage_options so that the default table storage orientation is column, and then using the ALTER TABLE command with the REORGANIZE=TRUE clause.

The issue occurred because the ALTER TABLE command with the REORGANIZE=TRUE clause did not recognize the correct table storage options during table data reorganization.

This issue has been resolved. The ALTER TABLE command detects the correct table storage options during table data reorganization.

29065 Query Execution 4.3.18.0 Greenplum Database does not support splitting a leaf child partition of a partition table if the leaf partition is defined as list partition. However, in some cases, attempting to split a list partition caused a Greenplum Database PANIC.

This issue has been resolved. Now Greenplum Database returns an error for the specified situation.

29061 gpcheckcat 4.3.18.0 In some cases, the Greenplum Database gpcheckat utility returned an error stating that a value was out of range for type integer when the utility did not handle large OID values correctly (relfile nodes exceeding 231, more than 2 billion).

This issue has been resolved. Now the utility handles large OID values correctly.

29046 Query Execution 4.3.18.0 For some group aggregate queries with a GROUP BY clause that contains a CUBE and also specify a HAVING clause, Greenplum Database generated a PANCIC. The PANIC was generated during aggregation when Greenplum Database did not handle a NULL case properly.

This issue has been resolved. For the specified type of queries, the handling of NULL cases during aggregation has been improved.

29012 gpload 4.3.18.0 In some cases, the Greenplum Database utility gpload recorded the incorrect number of rejected rows when loading data the when utility was configured to log errors. The utility did not properly record errors from multiple Greenplum Database segment instances.

This issue has been resolved. Now the utility correctly records errors from multiple segments.

29009 gprecoverseg 4.3.18.0 Running the Greenplum Database gprecoverseg with the option -r to rebalance segments did not return errors if the operation generated errors on some segments and completed successfully on other segments.

This issue has been resolved. Now if any errors occur during a rebalance operation, the user is notified and is referred to the logs for details.

26956 Query Optimizer 4.3.18.0 For some queries, GPORCA did not choose a query plan with bitmap index scan when a bitmap index on a column could be used as a filter condition in the query. This caused a performance degradation when comparing GPORCA with the legacy query planner.

This issue has been resolved. Now GPORCA chooses the faster plan with the bitmap index scan.

23438 gpcheckcat 4.3.18.0 The Greenplum Database gpcheckcat utility test missing_extraneous returned incorrect results when the test found database catalog inconsistencies. The test did not process the inconsistencies correctly.

This issue has been resolved. Now the utility returns correct results when the test finds catalog inconsistencies.

152083437 Query Execution 4.3.18.0 Some queries that perform in-memory operations generated an invalid memory alloc request error for some values of the statement_mem server configuration parameter. The error was caused when an in-memory operation exceeded a maximum allowed numeric value.

This issue has been resolved. The maximum allowed value has been increased to avoid the error.

150243109 Query Optimizer 4.3.18.0 GPORCA generated a PANIC when generating a plan for queries where the predicate performs a comparison of two subqueries, and the comparison also uses the clause ANY, ALL, IN, or NOT IN. GPORCA did not properly handle the subquery in left side of the comparison.

This issue has been resolved. Now GPORCA generates a plan successfully.

4.3.17.x Release

Table 16. Resolved Issues in 4.3.17.x
Issue Number Category Resolved In Description
29069 Query Optimizer 4.3.17.1 GPORCA query optimization performance was poor for queries that contain an IN operator with a large number of values and the values require an implicit CAST.

This issue has been resolved. Query optimization performance has been improved for the specified type of queries.

29060 Query Planner 4.3.17.1 When an UPDATE or DELETE command on a partitioned table contains a filter on partitioning key, the Greenplum Database legacy query optimizer produced a suboptimal plan. The plan contained a dummy join node with a One-Time False filter for each eliminated partition.

This issue has been resolved. Now the legacy optimizer does not create such dummy join nodes.

29042 Query Planner 4.3.17.1 The legacy query optimizer generated a Greenplum Database PANIC when executing some aggregation queries. The queries contain column aliases with the same name as the table columns, the queries contain subqueries that reference the column alias, and grouping is applied on the column alias. In some cases, the legacy optimizer caused a PANIC when it attempted to generate a plan that contained a inconsistent target list for the aggregation plan.

This issue has been resolved. The legacy optimizer generates the correct query plan for the specified type of aggregation queries.

29039 Query Optimizer 4.3.17.1 GPORCA generated a PANIC when a query attempted to use an index on a table column after the table was altered by dropping columns that are listed earlier in the table definition. For example, the index is defined on column 10, and columns 3 and 4 were deleted from the table. GPORCA did not correctly determine the index associated with the column.

This issue has been resolved. Now GPORCA correctly determines the column index in the specified situation.

29024 Query Planner 4.3.17.1

For some queries that contain a subquery, and the subquery contains a parameter referencing the outer query, the Greenplum legacy query optimizer generated an incorrect plan that did not handle the subplan correctly. A Greenplum Database PANIC occurred during query plan execution.

This issue has been resolved. The handling of subqueries by the legacy optimizer has been improved.

29005 Query Planner 4.3.17.1 For some DELETE or UPDATE commands that contain subqueries that reference the same table, the Greenplum Database legacy query optimizer generated a plan that did not correctly redistribute some rows to the appropriate segments. This caused a Greenplum Database PANIC when attempting to update or delete a row on an incorrect segment.

This issue has been resolved. Now the plan generated by the legacy optimizer correctly redistributes rows in the specified situation.

26993 gptransfer 4.3.17.1 The Greenplum Database utility gptransfer failed intermittently when transferring large amounts of data. When exporting data, Greenplum Database did not properly check whether a write operation had completed. This might cause intermittent failures when exporting large amounts of data in a single operation.

This issue has been resolved. Checking the completion of write operations during data export has been improved.

151498144 gpcrondump 4.3.17.1 When backing up and restoring database objects with the Greenplum Database utilities gpcrondump and gpdbrestore and specifying a schema level option such as the gpcrondump option -s or -S, functions (which are schema specific) were backed up and restored. However, only the PL/Java language was backed up and restored.

This issue has been resolved. Now for schema level back up and restore operations, procedural languages are included. See Changed Feature.

151344210 Query Planner 4.3.17.1 For an aggregation query, the Greenplum Database legacy query optimizer might return incorrect results when the aggregation is over a column, a CAST is defined on the column, and the original column (without the CAST) is in a GROUP BY clause. The legacy optimizer did not handle the CAST correctly in the query.

This issue has been resolved. Now the legacy optimizer handles the CAST correctly for the specified queries.

151341000 Backup and Restore 4.3.17.1 In some cases for a Greenplum Database system with 10 or more segment instances, the Greenplum Database gpdbrestore utility did not perform a restore operation correctly. When restoring from a version 4.3.11.3 or earlier backup to version 4.3.12.0 or later, the utility did not handle the backup filenames correctly. The format of backup filenames changed in 4.3.12.0.

This issue has been resolved. Now the utility, handles the backup filenames correctly in the specified situation.

151016845 Query Optimizer 4.3.17.1 For queries involving catalog tables, GPORCA falls back to Greenplum Database legacy optimizer. The fallback event was set to the log severity level LOG and was captured in the Greenplum Database logs. In some cases, the messages bloated the Greenplum Database log files. For example, when a workload executed a large number of queries against catalog tables.

The issue has been resolved. Now the log level for the fallback event is DEBUG1. This level does not log the fallback events by default. The log level can be set to log the events for debugging purposes.

150988530 Query Optimizer 4.3.17.1 In some cases, GPORCA generated incorrect results for a query when the query predicate contains a [NOT] EXISTS clause that contains both an aggregate function and a GROUP BY clause where a grouping column is a outer reference that is not in the aggregate function. GPORCA did not process the predicate correctly.

This issue has now been resolved. Now GPORCA handles the specified type of predicate correctly.

150906152 Query Optimizer 4.3.17.1 For some queries that contain a correlated subquery and the subquery contains a window function, GPORCA generated a query plan that returned incorrect results. The plan incorrectly performed the join after the window function was applied.

This issue has been resolved. Now the plan generated by GPORCA correctly applies the window function before the join.

29039 Query Optimizer 4.3.17.0 GPORCA generated a PANIC when a query attempted to use an index on a table column after the table was altered by dropping columns that are listed earlier in the table definition. For example, the index is defined on column 10, and columns 3 and 4 were deleted from the table. GPORCA did not correctly determine the index associated with the column.

This issue has been resolved. Now GPORCA correctly determines the column index in the specified situation.

29036 COPY 4.3.17.0 In some cases, a COPY command caused the Greenplum Database to go into recovery mode. This issue was caused by the format of metadata associated with data copied into the database.

This issue has been resolved. In the command, the checking of metadata has been enhanced.

29030 Query Optimizer 4.3.17.0 For queries that contain a large number of joins, GPORCA spent a significant amount of time evaluating all possible multiple join combinations to determine the most efficient join tree to use in the query plan.

This issue has been resolved. The maximum number of join combinations that are evaluated has been reduced.

29027 Backup and Restore 4.3.17.0 When performing a restore operation that specified a schema filtering option -S, the Greenplum Database utility gpdbrestore analyzed tables in all the schemas.

This issue has been resolved. Now the utility analyzes only the tables in the restored schemas.

29025 Backup and Restore 4.3.17.0 Some restore operations performed by the Greenplum Database utility gpdbrestore that used DD Boost failed because of issues parsing DD Boost path information in the backup report file.

This issue has been resolved. Now the utility correctly parses DD Boost path information.

29005 Query Planner 4.3.17.0 For some DELETE or UPDATE commands that contain subqueries that reference the same table, the Greenplum Database legacy query planner generated a plan that did not correctly redistribute some rows to the appropriate segments. This caused a Greenplum Database PANIC when attempting to update or delete a row on an incorrect segment.

This issue has been resolved. Now the plan generated by the legacy planner correctly redistributes rows in the specified situation.

28949 Query Planner 4.3.17.0 For some DELETE or UPDATE commands that contain subqueries that reference the same table, the Greenplum Database legacy query planner generated a plan that caused a Greenplum Database PANIC. The plan did not account for some columns projected by the join.

This issue has been resolved. Now the plan generated by the legacy planner accounts for all columns in the specified situation.

27011 gpload 4.3.17.0 On AIX 7.1 systems, the Greenplum Database client loader utility gpload failed with a Python not found message.

This issue has been resolved. The Greenplum Database Client Loader tools have been updated.

26834 Transaction Management 4.3.17.0 A DDL or VACUUM operation that creates or drops an append-optimized segment file requires a shared buffer. In some cases, this operation caused a deadlock with other operations that require the shared buffer.

This issue has been resolved. To avoid a deadlock in the specified situation, the handling of locks for the shared buffer has been improved.

26832 gpcheckcat 4.3.17.0 When the Greenplum Database utility gpcheckcat checked multiple databases in a single operation, the utility incorrectly reported catalog issues for a database when the issue was found in a database that was checked earlier.

This issue has been resolved. Now when checking multiple databases, the utility reports the catalog issues correctly for each database.

26825 Query Planner 4.3.17.0 For queries that contain an IN clause that includes a correlated subquery with a LIMIT or OFFSET clause, the legacy query planner did not properly handle the LIMIT or OFFSET clause. This issue generated a SIGSEGV.

This issue has been resolved. Now the query planner handles the specified type of queries correctly.

26778 gpinitsystem 4.3.17.0 The Greenplum Database utility gpinitsystem did not initialize a Greenplum Database system correctly when the initialization of the standby master failed. The standby master failure caused errors creating gp_toolkit administrative schema and setting server configuration parameters. The failure and errors were not handled correctly.

These issues have been resolved. Now the failure of the standby master initialization, does cause the gp_toolkit and parameter issues. Also, error handling and logging of the specified errors have been improved.

26598 gpload 4.3.17.0 The Greenplum Database utility gpload would intermittently hang. The issue was caused by gpfdist instances that were used by gpload. A function used by gpfdist was not re-entrant.

This issue has been resolved. Now the function is re-entrant.

26363 GPHDFS 4.3.17.0 Performance was poor when reading data from Parquet files with a Greenplum Database external table defined with the gphdfs protocol.

This issue has been resolved. Performance has been improved when reading Parquet data with the gphdfs protocol.

15254 gpconfig 4.3.17.0 The Greenplum Database utility gpconfig did not properly set the server configuration parameter pljava_classpath. After setting the parameter, restarting Greenplum Database with the gpstop -arcommand would fail. Specifying the value for the parameter in the command required enclosing the value in single quotes and then in double quotes.

This issue has been resolved. The use of quotes when specifying parameter values has been simplified.

151028970 Query Optimizer 4.3.17.0 For queries that contain a FULL OUTER JOIN with join condition ON FALSE, GPORCA generated an incorrect plan that returned wrong results. GPORCA incorrectly returned an empty set for the full outer join that caused the query plan to return no tuples.

This issue has now been resolved. Now GPORCA generates a correct plan.

149524459 Query Optimizer 4.3.17.0

When GPORCA attempted to CAST an integer array to a double precision array, GPORCA did not perform the CAST of the array elements correctly.

This issue has been resolved. GPORCA now performs the CAST correctly.

4.3.16.x Release

Table 17. Resolved Issues in 4.3.16.x
Issue Number Category Resolved In Description
28977 analyzedb 4.3.16.1 When the Greenplum Database utility analyzedb ran multiple, concurrent analyze operations from different analyzedb instances, the analyzedb snapshot information was not updated correctly. The snapshot information contains information about analyzedb operations. This might have caused analyzedb to incorrectly skip analyzing tables.

This issue has been fixed. The management of analyzedb snapshot information has been improved.

26910 Query Optimizer 4.3.16.1 In some cases, GPORCA caused a Greenplum Database PANIC when running a query after cancelling a query that was displaying informational log messages. The cancelled query caused an inconsistent memory pool state that caused a PANIC when the subsequent query was run.

This issue has been resolved. GPORCA interrupt handing has been improved for the specified situation.

26701 gpmigrate 4.3.16.1 When running the Greenplum Database utility gppkg with the --migrate option, the utility did not correctly migrate the Greenplum Database extensions installed in an existing installation to an upgrade installation.

This issue has been resolved. Running gppkg with the --migrate option migrates installed extensions. Now, at a minimum, the master must be running on the upgrade Greenplum Database system. See Changed Features.

149956098

gpload 4.3.16.1 When running the Greenplum Database utility gpload, the size of log files increased quickly in some situations. The logs contained information from the gpfdist instances that were used by gpload. This caused the gpload logs to increase quickly in some cases.

This issue has been resolved. The gpload utility no longer includes the gpfdist information to the log files as the default. The gpload option -v or -V must be specified to include gpfdist information.

148821441 PL/Java 4.3.16.1 After installing the PL/Java extension package with the Greenplum Database utility gppkg, running a Java application returned a Java error that stated that libjvm.so did not exist. The error was caused by the JAVA_HOME environment variable that was incorrectly set in the greenplum_path.sh file.

This issue has been resolved. The PL/Java extension package has been updated to set JAVA_HOME correctly. The new PL/Java package version is pv1.3.1. See Changed Features.

28989 Backup and Restore 4.3.16.0 When the Greenplum Database gpdbrestore utility attempted to restore a schema from a full backup by specifying a schema filtering option -S, the restore operation failed if a user-defined function contains a BEGIN ... END block and the function is not qualified by the schema. The utility would incorrectly run the BEGIN ... END block.

This issue has been resolved. Now the utility correctly handles user-defined functions in the specified situation.

28972 Query Optimizer 4.3.16.0 GPORCA generated a Greenplum Database PANIC when a query contains a function where an input argument is a subquery. GPORCA did not handle the subquery correctly and generated a PANIC.

This issue has been resolved. Now GPORCA correctly processes subqueries when they are input arguments to a function in a query.

28968 Backup and Restore 4.3.16.0 When running the Greenplum Database gpdbrestore utility with the -S option to restore database objects qualified by a specified schema from a backup, the utility was not restoring access privilege information (GRANT privileges) for the schema.

This issue has been resolved. When restoring database objects with the -S option, the access privilege information for the schema is restored.

28967 Backup and Restore 4.3.16.0 When restoring a database with the gpdbrestore utility, the access privilege information (GRANT privileges) that was set for the database was missing. The missing access privilege information was not being backed up by the Greenplum Database gpcrondump utility.

This issue has been resolved. Now, the gpcrondump utility backs up the database level access privilege information and is restored by the gpdbrestore utility.

28963 Query Executor 4.3.16.0 In some cases, when Greeplum Database executes a query with a JOIN, when the hash side of the hash join operator is empty, an optimization in the executor skips executing the probe side of the join operator ("squelching"). Previously the executor was too aggressive in squelching. If the probe side of the hash join has SharedInputScans, there could be circumstances where a SharedInputScan reader could hang if the SharedInputScan writer was squelched.

This issue has now been resolved in the Greeplum Database executor.

27007

26964

Query Planner 4.3.16.0 For queries that contained distinct qualified aggregates (DQA) and grouping sets (GROUPING SET, ROLLUP, CUBE), the Greenplum Database legacy optimizer incorrectly replaced the grouping columns appearing in the projection list while producing a plan with an Append node. This caused a Greenplum Database PANIC.

This issue has been resolved. Now the legacy optimizer generates a correct plan for the specified queries.

27000 Transaction Management 4.3.16.0 In some cases, Greenplum Database issued an error indicating a missing chunk number for a toast value. The error could have occurred when a mirror segment took over as a primary segment or when a rebalance occurred. The error was caused by to the improper handling of blocks of data during incremental recovery with the Greenplum Database gprecoverseg utility. The improper data handling caused the mirror segment to be incorrectly marked as in-sync with primary segment.

This issue has been resolved. The handling of blocks during incremental recovery has been improved.

26908 GPHDFS 4.3.16.0 Greenplum Database performance was poor for queries with a LIMIT clause that accessed an external table with the gphdfs protocol. The poor performance was caused by Greenplum Database not efficiently managing the freeing of resources after the query completes.

This issue has been resolved. Greenplum Database has improved how resources are freed for queries against external tables.

26865 gpfdist 4.3.16.0 When an HTTP request with a malformed or root (/) URL was sent to the Greenplum Database utility gpfidst, the utility resets and generates an error that extra data was detected .

This issue has been resolved. The utility has been enhanced and now detects issues with a HTTP request and returns an error message indicating the error.

26663 Locking, Signals, Processes 4.3.16.0 In some cases on a Greenplum Database system, a running query could not be canceled with the pg_cancel_backend() command. The query could not be cancelled because during query processing, a backend query process was placed in a state that did not check for an interrupt such as a pg_cancel_backend() command.

This issue has been resolved. Checking for interrupts by backend query processes have been improved.

26441 gpfdist 4.3.16.0 For a Greenplum Database writable external table defined with the gpfdists protocol (gpfdist with SSL) was much slower that a table defined with the gpfdist protocol.

This issue has been resolved. The gpfdists protocol performance has been improved by upgrading the OpenSSL version used by Greenplum Database.

148133765 PL/R 4.3.16.0 After installing the Greenplum Database PL/R extension package version pv2.2 on a CentOS 7.x or RHEL 7.x system, running the rpm utility failed. with the message version `XZ_5.1.2alpha' not found (required by /lib64/librpmio.so.3) This issue also affects the Greenplum Database gppkg utility that uses rpm.

This issue has been resolved. Installing the PL/R extension package version pv2.3 does not cause the rpm failure.

147931843 Security 4.3.16.0 Greenplum Database software that runs on Linux systems has been updated to use OpenSSL 1.0.2l. For information about major changes in OpenSSL 1.0.2l, see https://www.openssl.org/news/openssl-1.0.2-notes.html

4.3.15.0 Release

Table 18. Resolved Issues in 4.3.15.0
Issue Number Category Resolved In Description
27961 Backup and Restore 4.3.15.0 In some cases, the Greenplum Database utility gpdbrestore failed when specifying a filter option such as -T or --table-file. The utility failed if a table that is being restored contains an XML data column, and the column contained data that is a valid XML content (XML fragment) but is not an XML document.

This issue has been resolved. The utility restores the XML data for the specified situation.

27035 Dispatch 4.3.15.0 In some cases when a SIGINT signal is issued during query execution, a Greenplum Database query dispatcher (QD) did not correctly terminate a Greenplum Database gang (a set of query processes that are working on a segment). This caused Greenplum Database to respond slowly or become non-responsive.

This issue has been resolved. QD gang management has been improved to terminate gangs in the specified situation.

27024 Backup and Restore 4.3.15.0 The Greenplum Database utility gpcronump did not correctly back up ordered aggregate functions.

This issue has been resolved. Now, ordered aggregate functions are backed up correctly.

26926 Query Planner 4.3.15.0 For some common table expression (CTE) queries, queries that contain a WITH clause, the legacy query optimizer did not handle the WITH subquery correctly if the subquery contains a predicate with an EXISTS clause. Incorrect results were returned.

This issue has been resolved. The legacy optimizer handling of the specified type of CTE queries has been improved. See Product Enhancements.

26911 Backup and Restore 4.3.15.0 In some cases the Greenplum Database utility gpdbrestore did not correctly restore table statistics when a table and the table statistics were restored from a backup. The statistics were not restored correctly if a column was dropped from the table before the table was backed up. The incorrect table statistics might cause queries against the table to fail.

This issue has been resolved. The statistics are restored correctly for the specified situation.

26901 gpload 4.3.15.0 When the Greenplum Database utility gpload was running and the utility was stopped, gpfdist instances that were used by gpload were not stopped.

This issue has been resolved. Now, the gpfdist instances that are used by gpload are also stopped.

26785 gpssh 4.3.15.0 In some cases, the Greenplum Database gpssh utility, could not connect to a remote Greenplum host system due to network timeout issues.

This issue has been resolved. The utility supports the ability to attempt connecting to a host multiple times before returning an error. See Product Enhancements.

26773 Backup and Restore 4.3.15.0 When running the Greenplum Database gpdbrestore utility with the --dump-stats option to restore database table statistics, and statistics already exits for a table column, statistics would not be correctly restored.

This issue has been resolved. Now, before a gpdbrestore restore operation restores statistics, existing statistics are deleted.

26437 Backup and Restore 4.3.15.0 Running the Greenplum Database utility gpcronump with the option --list-backup-file that creates two text files that contain the names of backup files that gpcrondump will create when backing up a database. The name of the backup file that contains the database statistics was not included in the list of files that be can be created as named pipes when the --dump-stats option was also specified.

This issue has been resolved. When the --dump-stats option is specified with the --list-backup-file option, the name of the statistics backup file is listed.

148098229 Query Optimizer 4.3.15.0 For queries that contain a filter of the form CONST = VAR (for example, 2 = myvar), GPORCA used the incorrect selectivity to estimate the cardinality. This could result in GPORCA creating suboptimal plans.

This issue has been resolved. GPORCA uses the correct cardinality estimate for the specified type of filter.

147709955 Query Parser 4.3.15.0 If a view definition contains COUNT(*) with a FILTER condition, Greenplum Database did not store the text of the view definition correctly. The view definition that is executed was stored correctly. If a view definition contains COUNT(*) with a FILTER condition, the Greenplum Database query parser converts the phrase into a COUNT of a CASE - WHEN expression. The expression was incorrectly omitted from view definition text.

This issue has been resolved. Now, the text of the view definition is stored correctly.

147706253 Query Planner 4.3.15.0 In some cases, the Greenplum Database legacy query optimizer returned incorrect results when the query contains the boolean comparison IS NOT FALSE. The legacy optimizer incorrectly considered expression IS NOT FALSE the same as expression (the expression is true). When a NULL is returned as a boolean value, for example 't'::boolean = NULL, the IS NOT FALSE comparison handles the value differently than the default comparison (the value is true).

This issue has been resolved. Now, the legacy planner optimizer the IS NOT FALSE comparison correctly.

147631003 Backup and Restore 4.3.15.0 When the Greenplum Database utility gpdbrestore was run with --restore-stats option to perform a restore operation from a DDBoost system, statistics were not restored.

This issue has been resolved. Database statistics are restored in the specified situation.

147455519 gptransfer 4.3.15.0 In some cases when the Greenplum Database utility gptransfer copied tables between databases, the table column names were not being processed correctly and caused temporary intermediate external tables to be created incorrectly. The table was being created correctly in the destination database, but caused gptransfer to incorrectly display warning messages.

This issue has been resolved. Now, the external tables are created correctly and no warnings are generated.

146718261 gpperfmon 4.3.15.0 In some cases, when gpperfmon attempted to collect skew metrics for a query process ID (PID) that had already exited, the correct skew metric values could be overwritten with a value of 0.

This issue has been resolved. Metric validation has been added to ensure the correct skew metrics are collected.

146655909 Backup and Restore 4.3.15.0 In some cases, the Greenplum Database utility gpcrondump did not correctly back up the URI in the LOCATION clause of an external table definition. The utility incorrectly enclosed the URI in quotes when there is a space in the URI string.

This issue has been resolved. Now, the external table definition is backed up correctly.

4.3.14.x Release

Table 19. Resolved Issues in 4.3.14.x
Issue Number Category Resolved In Description
CVE-2017-7485 Security 4.3.14.1 Fixes a potential SSL downgrade issue in libpq. The PGREQUIRESSL environment variable was no longer enforcing a SSL/TLS connection to a PostgreSQL server. A man-in-the-middle attack could use this flaw to strip the SSL/TLS protection from a connection between a client and a server. See CVE-2017-7485.

This security vulnerability has been patched.

26990 gphdfs 4.3.14.1 When attempting to access external table data with the gphdfs protocol, Greenplum Database returned an error stating that the file hadoop_env.sh could not be found. The gphdfs protocol could not access the external table data because it did not search the correct path for the hadoop_env.sh file.

This issue has been resolved. The gphdfs protocol now searches correct path for the file.

146655909 gpcrondump 4.3.14.1 The Greenplum Database utility gpcrondump did not correctly handle the quotes that are in the LOCATION clause of the CREATE EXTERNAL TABLE command.

This issue has been resolved. Now the quotes in the LOCATION clause are handled correctly.

146011767 gpexpand 4.3.14.1 In some cases, when the Greenplum Database utility gpexpand was run with a specific duration or end-time by specifying the option -d or -e, the utility incorrectly reported a that a table redistribution was COMPLETED when it was not. This also caused the table distribution policy to be changed to distributed randomly.

This issue has been resolved. For the specified situation, the gpexpand utility correctly reports the table redistribution policy and reverts the table redistribution.

26947 Query Optimizer 4.3.14.0 When GPORCA generated a plan for a query with a correlated nested loop join (NLJ), GPORCA incorrectly discarded a subplan if the plan contained an outer child node as partition selector and an inner child node as a subplan that is an additional filter condition. This caused the query to return incorrect results.

This issue has been resolved. The handling of subplans for correlated NLJ has been improved.

26941

26252

Query Planner 4.3.14.0 In Greenplum Database, some queries with a complex join condition with these characteristics failed with an interconnect error during execution:
  • The join conditions are combined with an AND clause
  • Multiple join predicates on the same table column, with some predicates containing a subquery and others comparing the column to other columns.

For the specified type of query, the legacy planner inferred new predicates from the simple join predicate and the predicate with a subquery. The planner failed to properly apply motions to the subquery because the same subquery was being referred to by two or more predicates. The resulting plan failed during execution.

This issue has been resolved. The handling of specified type of query has been improved.

26917 Interconnect 4.3.14.0 In a transaction, when the server configuration parameter gp_interconnect_queue_depth was changed after a cursor was declared, a Greenplum Database PANIC might be generated. The PANIC was caused by improper handling of the parameter.

This issue has been resolved. The handling of the parameter within a transaction has been improved.

26884 Catalog and Metadata 4.3.14.0 In some cases, performing ALTER TABLE ADD COLUMN operations on an append-optimized column-oriented table did not update the table metadata correctly. This caused some database catalog inconsistencies and prevented the VACUUM command from recovering space from the table.

This issue has been resolved. Handling of append-optimized column-oriented table metadata has been improved.

26855 Query Execution 4.3.14.0 In some cases, Greenplum Database incorrectly returned no results for a query that contains a WITH clause, a common table expression (CTE). During query execution, Greenplum Database did not correctly process shared input scans and redistribute motions that were in the same slice.

This issue has been resolved. Now the operations are processed correctly for the specified situation.

26842

144497579

Scripts: Backup and Restore 4.3.14.0 During a restore operation with the Greenplum Database utility gpdbrestore, the operation failed if a table is defined with an XML data column, and the column contained data that is a valid XML content (XML fragment) but is not an XML document. The failure is caused when the COPY command that is used by the utility issues a error when the command tries to copy the XML content fragment.

This issue has been resolved. When backing up data, Greenplum Database now sets the server configuration parameter gp_strict_xml_parse to false. Setting the value to false allows the gpdbrestore utility to restore the XML data. See the server configuration parameter gp_strict_xml_parse in the Greenplum Database Reference Guide.

IMPORTANT: The parameter is set only for new backups created by Greenplum Database.

26770 Storage: Transaction Management 4.3.14.0 In some cases, Greenplum Database generated an error that states that a persistent tuple is greater than the last known TID. The error was caused by improper locking of shared memory during data processing of persistent data.

This issue has been resolved. Locking shared memory in the specified situation has been improved.

145169165 gpperfmon 4.3.14.0 When the Greenplum Database utility gpperfmon_install was run to install the gpperfmon database and configure a Greenplum Database system for performance monitoring, the utility did not configure the system to connect to the gpperfmon database with IPv6.

This issue has been resolved. The utility now configures a Greenplum Database system accept gpmon user connections over IPv6.

143543637 Query Execution 4.3.14.0 For queries with subqueries, the Greenplum Database legacy query optimizer could generate a plan that incorrectly removed an InitPlan node. This caused a query executor error.

This issue has been resolved. The legacy planner has been enhanced to evaluate InitPlan nodes correctly for the specified type of queries.

143542727 Query Execution 4.3.14.0 The Greenplum Database query executer was returning an error or crashing for some queries against tables with columns that are defined with a bitmap and btree index, and the query specified the columns in the WHERE clause. The issue was caused by an attempt to free memory that was already freed.

This issue has been resolved now. Memory management has been enhanced for the specified queries.

143221115 gpperfmon 4.3.14.0 Queries that were recorded in the gpperfmon database had line breaks removed. This caused issues with readability and potentially query parsing. The gpperfmon database is a dedicated database where monitoring agents on Greenplum Database segment hosts save statistics.

This issue has been resolved. Line breaks are no longer removed when storing queries.

141804111 Scripts: Backup and Restore 4.3.14.0 In some cases, running the Greenplum Database utility gpdbrestore -s dbname fails when Greenplum Database contains more than one backup directory with a date stamp YYYYMMDD as the directory name in the db_dumps directory. The failure occurs when the utility searches for the latest timestamp report file in the wrong directory.

This issue has been resolved. Now the utility searches the correct directory.

137815123 VACUUM 4.3.14.0 In some cases when concurrent transactions executed a VACUUM command on different append-optimized tables, the age of some of the tables (and their auxiliary tables) was not decreased. The age was not decreased because multiple concurrent VACUUM process were not handled properly.

This problem is resolved. Now, multiple concurrent VACUUM processes are handled properly for the specified situations.

4.3.13.0 Release

Table 20. Resolved Issues in 4.3.13.0
Issue Number Category Resolved In Description
26901

26865

26712

26371

gpfdist, gpload 4.3.13.0 In some cases, the Greenplum Database utility gpfdist (and utility gpload that uses gpfdist) caused various issues. For example, the utility failed intermittently, did not complete transferring data, and remained running after a database shutdown. These issues were caused by library that is used by gpfdist.

These issues have been resolved. The library has been updated to a version that resolves the issues.

26881 Storage: Segment Mirroring 4.3.13.0 In some cases, running the Greenplum Database utility gprecoverseg -r to rebalance segments that were marked as down generated a FATAL error stating that a file was missing. The error was caused due to the improper handling of transactions during the recovery process.

This issue has been resolved. Transaction management has been improved during the recovery process.

26839 Storage: Filespace / Tablespace 4.3.13.0 In some cases, queries against a column-oriented table defined with an index returned an expected header error. The error was caused by the improper generation of an index when the size of data in the indexed column is large.

This issue has been resolved. The index is generated correctly in the specified situation.

26786 Scripts: Backup and Restore 4.3.13.0 A backup operation performed by the Greenplum Database utility gpcrondump to a Data Domain system failed (the --ddboost option is specified) if the database does not contain append-optimized (AO) or column-oriented (CO) tables. The failure was caused when the utility backed up empty (0 byte) AO or CO state files.

This issue has been resolved. The utility has improved how the empty files are handled. As a workaround for existing backups, a return can be added to the empty file.

26753 Query Optimizer, Query Planner 4.3.13.0 In some cases, the legacy query optimizer generated a Greenplum Database PANIC when the legacy optimizer encountered corrupted statistics during query optimization.

This issue has been resolved. Now, when the legacy query optimizer detects corrupted statistics, the corrupted statistics are skipped and a warning is issued.

26707 Scripts: Backup and Restore 4.3.13.0 When performing a restore operation with the Greenplum Database gpcrondump utility that specifies a table filtering option such as --table-file, the restore operation did not restore tables that are defined with zero columns. A full restore operation restores tables that are defined with zero columns.

This issue has been resolved. The utility restores tables that are defined with zero columns in the specified situation.

26637 Query Optimizer 4.3.13.0 GPORCA query optimization performance was poor for queries that contain an IN operator with a large number of values.

This issue has been resolved. Query optimization performance has been improved for the specified type of queries. See Product Enhancements.

26511 Scripts: Backup and Restore 4.3.13.0 In some cases, when running the Greenplum Database utility gpdbrestore with the --ddboost option, the restore operation failed. The operation failed due to improper file handling that caused the maximum open file limit being reached.

This issue has been resolved. File handling has been improved to properly close files in the specified situation.

26441 gpfdist 4.3.13.0 In some cases, performance was poor when downloading data to an external table when the Greenplum Database utility gpfdistwas used with SSL encryption (with the --ssl option).

This issue has been resolved. The performance of he gpfdist utility has been enhanced when the --ssl option is specified. See Product Enhancements.

142566503 Query Optimizer 4.3.13.0 For queries against partitioned tables, GPORCA did not correctly compare partitions during query optimization in some situations. This incorrect comparison increased optimization times.

This issue has been resolved. Now, GPORCA correctly compares partitions that were previously compared incorrectly.

141890639 Scripts: gpstart 4.3.13.0 When running the Greenplum Database gpstart utility with the -m and -a options, the utility improperly displayed the confirmation prompt. The options are specified to start only the Greenplum Database master instance (for maintenance tasks) without prompting for confirmation.

This issue has been resolved. The utility does not display prompts in the specified situation.

141826453 Query Optimizer 4.3.13.0 GPORCA generated an incorrect query plan for some queries against a multi-level partitioned table that is defined with a default partition. Incorrect plans were generated if the query contains equality predicates on the table, and one or more partition levels do not have an equality predicate on the partition key. The plan incorrectly selected the default partition as well as the desired partitions.

This issue has been fixed. GPORCA selects the correct partitions in the specified situation.

141342591 Scripts: gpstop 4.3.13.0 When running the Greenplum Database gpstop utility with the -m option to stop a Greenplum Database master instance that was started in maintenance mode, the utility did not prompt the user for confirmation.

This issue has been resolved. The utility prompts the user for confirmation. If the -a option also specified, the confirmation prompt is not displayed.

140601033 Query Optimizer 4.3.13.0 For queries with a large number of constrains, GPORCA query optimization performance was poor due to the method of evaluating the query constraints and determining equivalence classes.

This issue has been resolved. The method of determining equivalence classes has been improved. See Product Enhancements.

4.3.12.0 Release

Table 21. Resolved Issues in 4.3.12.0
Issue Number Category Resolved In Description
26798 Dispatch 4.3.12.0 In some cases, when executing a query that contains an exception block in a function, Greenplum Database might have caused a SIGBUS or SIGSEGV error. The error occurred because Greenplum Database incorrectly dispatched a command to a busy gang process on a segment when executing the exception block.

This issue has been resolved. Now, Greenplum Database aborts the query and prints an error message in the specified situation.

26790 Scripts: gpcheckcat 4.3.12.0 When the Greenplum Database gpcheckcat utility completed, the completion message displayed the incorrect time.

This issue has been resolved. Now, the utility displays the correct time.

26787 Dispatch 4.3.12.0 In some cases, a Greenplum Database PANIC occurred when a query that accessed an external table failed. Greenplum Database did not correctly handle an error that occurred when a scan of a external table failed.

The issue has been resolved. Now, the specified error is handled properly.

26782 Query Optimizer 4.3.12.0 For queries that require a full join on append-optimized partitioned tables, GPORCA did not handle non-visible system columns on the tables correctly. This generated an incorrect plan that caused an error in the query executor.

This issue has been resolved. Now, GPORCA handles non-visible system columns correctly in the specified situation.

26760 Query Optimizer 4.3.12.0 For queries with a projection list that contains multiple set returning functions, GPORCA incorrectly collapsed the projected nodes. This caused GPORCA to return incorrect results.

This issue has been resolved. GPORCA correctly handles the specified type of queries.

26753 Query Planner 4.3.12.0 In some cases, the legacy query optimizer generated a Greenplum Database PANIC when the legacy optimizer encountered corrupted statistics during query optimization.

This issue has been resolved. Now, when legacy query optimizer detects corrupted statistics, the corrupted statistics are skipped and a warning is issued.

26719 DML, Storage: Access Methods 4.3.12.0 For a column-oriented table, in some rare cases, data might not be stored correctly in a column when the column is defined with data type BIGINT and with RLE compression. The error occurred due to the incorrect handling of data during the compression process..

This issue has been resolved. The data compression process has been improved.

26706 Query Execution 4.3.12.0 For some queries, the query plan generated by GPORCA did not correctly handle bitmap index information during query execution. This caused a Greenplum Database PANIC.

The issue has been resolved. Handling of bitmap index information during query execution has been improved.

26573 Storage: Vacuum/Reindex/Truncate 4.3.12.0 In some cases, when a lazy vacuum operation was performed by the VACUUM command, Greenplum Database did not perform a compaction of the database segment files when ratio of hidden rows to total rows in the segment file was greater than the Greenplum Database server configuration parameter gp_appendonly_compaction_threshold. Compaction was not performed because the ratio was not calculated correctly.

This issue has been resolved. The ratio calculation has been improved. For information about the parameter, see the Greenplum Database Reference Guide.

26440 Scripts: gpconfig 4.3.12.0 The server configuration parameter value displayed by the Greenplum Database gpconfig utility with the -s option is the value from the database, not the value from the master and segment instance postgresql.conf files.

The parameter value from the postgresql.conf files can be displayed by specifying the gpcoonfig option --file with the -s option. See Product Enhancements.

20620 DDL 4.3.12.0 The ALTER TABLE RENAME command did not acquire an Access Exclusive lock on the table.

This issue has been resolved. See Changed Features.

136986803 Query Execution 4.3.12.0

Greenplum Database returned incorrect results for a query when the legacy query planner generated query plan that contained Merge Full Join and both children of the join are Motion operators. Incorrect results were returned because not all data was returned by a Motion operator.

This issue has been resolved. Now Motion operators return all the data in the specified situations.

n/a Server Configuration 4.3.12.0 The server configuration parameter pljava_classpath_insecure, described in earlier versions of the 4.3.11.x documentation, is not available in this Greenplum Database release.

This issue has been resolved. See New Parameter.

4.3.11.x Releases

Table 22. Resolved Issues in 4.3.11.x
Issue Number Category Resolved In Description
26750 Scripts: gptransfer 4.3.11.3 The Greenplum utility gptransfer did not correctly copy tables that are defined with multiple distribution keys.

This issue has been resolved. Now, the utility copies tables that are defined with multiple distribution keys. See Changed Features.

26744 Query Planner 4.3.11.3 In Greenplum Database, the legacy query planner might have generated a plan that returned incorrect results for window queries that have a subquery that contains a table valued function, and the function contains a non-correlated subquery.

This issue has been resolved. Now, the query planner generates a plan that returns correct results for the specified type of query.

137576977 Scripts: gpstop 4.3.11.3 On a Greenplum Database system with an active standby master, running the Greenplum Database utility gpstop with the default -M option, -M smart, caused a long delay before Greenplum Database shut down.

This issue has been resolved. The unnecessary delay has been removed.

26716 Dispatch, Scripts: recoverseg 4.3.11.2 In some cases, running the Greenplum Database gprecoverseg with the -r option did not complete (hung), did not return primary and mirror segments to their preferred roles, and generated a PANIC on the Greenplum Database master.

The issue was caused by the improper handling of process gangs during termination (processes that are running on different segments that were working on the same slice of a query plan).

This issue as been resolved. Gang termination handling has been improved.
26714 Query Optimizer 4.3.11.2 In some cases when Pivotal Query Optimizer (PQO) was enabled, PQO generated a sub-optimal query plan that broadcasts a relation with large cardinality that exhausted Greenplum Database resources during execution with a message stating that the workfile per query size limit was exceeded

This issue has been resolved. PQO has been enhanced to avoid a broadcast motion in query plans when a relation has a large cardinality.

26705

26657

Query Planner 4.3.11.2 For some queries that contain a correlated subquery, the Greenplum Database legacy query planner generated incorrect results or caused a Greenplum Database PANIC. The issue occurred when the subquery contained a binary operator and both operator expressions refer to outer variables. The query planner incorrectly assumed that one operator expressions is referring to an inner variable.

This issue as been resolved. Now for the specified type of queries, the legacy query planner generates the correct plan that properly evaluates the operator expressions.

26703 Scripts: gpstart/gpstop 4.3.11.2 In some cases, the Greenplum Database gpstart utility failed after a server reboot because a startup task did not complete within a specific time window. A subsequent restart completed successfully.

This issue has been resolved. The gpstart utility has been enhanced to account for the specified situation.

26695 Scripts: analyzedb 4.3.11.2 In some cases, the performance was poor when running the Greenplum Database analyzedb utility.

Performance has been improved. The value of the server configuration parameter statement_mem to 125MB for the database session that performs the analyze operations. See Changed Features.

26694 Scripts: Backup and Restore 4.3.11.2 Performance was poor when backing up a large database with the Greenplum Database gpcrondumputility with the --dump-stats option.

Performance has been significantly improved by caching the connection to the database instead of opening new connections when backing up database statistics. See Changed Features.

26691 Analyze 4.3.11.2 An ANALYZE performance regression was caused by a fix to the Greenplum Database legacy query planner that requires sending type information to the segments to avoid a system crash.

The ANALYZE performance issue has been resolved.

26644 Query Planner 4.3.11.2 The Greenplum Database legacy query planner generated a PANIC when generating the query plan if the query contains the GROUP BY clause with a ROLLUP() expression that contains a column that is a constant value and the same constant column is in the projection list.

This issue has been resolved.

26592 Query Optimizer 4.3.11.2 When Pivotal Query Optimizer (PQO) was enabled, data loaded from an external table to randomly distributed table (table definition contains the DISTRIBUTED RANDOMLY clause) was loaded onto a single segment instance.

PQO has been enhanced to redistribute rows across all segment instances when loading data to a randomly distributed table. The enhancement also improves the performance of parallel loading from an external table.

133058033 Scripts: gptransfer 4.3.11.2 The Greenplum Database gptransfer utility failed to validate a table when the table name contains capital letters.

This issue as been resolved. Now, the utility validates table names that contain capital letters.

26735 Upgrade/Downgrade 4.3.11.1 Upgrading from Greenplum Database 4.3.x.x to Greenplum Database 4.3.11.0 failed due to a persistent table issue. The issue occurred if a database was dropped from the Greenplum Database system prior to the upgrade.

The issue has been resolved in Greenplum Database 4.3.11.1.

Warning: Greenplum Database 4.3.11.0 was removed from General Availability. Do not upgrade to Greenplum Database 4.3.11.0 and do not attempt to change an existing system with data from 4.3.11.0 to another version.
26688 Query Execution 4.3.11.0 The following types of queries failed and returned a message stating that no such file or directory exists: queries with a cursor statement or some queries from a JDBC or ODBC driver. The failures occurred when there were utility mode connections to Greenplum Database segment instances. The failures occurred because the Greenplum Database did not properly track the utility mode connections.

Now, Greenplum Database correctly tracks utility mode connections to segment instances.

26669 Query Optimizer 4.3.11.0 Pivotal Query Optimizer (PQO) did not correctly rewrite some queries that contain an outer reference in a GROUP BY clause and did not use the outer reference in another aggregate function. The incorrect rewire caused a Greenplum Database PANIC.

Now, PQO performs additional validations to ensure the specified type of query is correctly rewritten.

26665 Transaction Management 4.3.11.0

In some cases, Greenplum Database did not properly handle SQL transactions that consist of multiple SQL statements with a COMMIT or END statement that occurs before a statement that requires a global subtransaction. This issue might force the database into recovery mode, or might cause a Greenplum Database PANIC.

The processing of global subtransactions for the specified type of SQL transactions has been improved.

26659 Scripts: Backup and Restore 4.3.11.0 In some cases, the Greenplum Database gpcrondump utility failed when performing an incremental backup operation to a Data Domain system with the --ddboost option. The backup failed when the report and cdatabase files were generated on different days (for example, if the backup is started just before midnight).

Now, the gpcrondump utility uses the backup timestamp instead of the current timestamp when generating files to avoid this issue.

26654 Dispatch 4.3.11.0 In some cases during Greenplum Database query execution, communication between the query dispatcher (QD) and a query executor (QE) was not handled properly. This caused the QE to not recognize a cancel request and caused the query execution to hang.

Message validation between QD and QE has been improved.

26631 Query Optimizer 4.3.11.0 Pivotal Query Optimizer (PQO) generated an incorrect cardinality estimate for queries that contain an OR predicate with these properties.
  • Both conditions in the OR expression are on the same column.
  • One condition of the OR expression uses <> (not equal) or NOT IN, the other condition is an IS NULL condition.

Incorrect estimates can prevent PQO from generating an optimal plan.

PQO cardinality estimation has been enhanced.

26613 Query Optimizer 4.3.11.0 For some Pivotal Query Optimizer (PQO) errors, PQO incorrectly returned a WARNING message followed by an ERROR message.

Now, PQO returns only the ERROR message.

26612

26600

26576
Catalog and Metadata 4.3.11.0 In some cases, a PANIC on a Greenplum Database segment was caused by inconsistencies in a system catalog table. The inconsistencies were caused by incorrect data being inserted into the table.

Additional checks have been added to Greenplum Database that prevent incorrect data from being inserted into system catalog tables.

26605 Locking, Signals, Processes 4.3.11.0 In some cases while processing append-optimized table tables, Greenplum Database returned an invalid page header message that did not specify the correct relation ID.

Now, the message includes the correct relation ID.

26603 Query Planner 4.3.11.0 In Greenplum Database, for queries that contain a correlated EXIST subquery and a WITH clause, the legacy planner generated a correlated plan that incorrectly attempted to access parameters across slices. This caused the query executor (QE) to panic.

Now, the legacy query planner generates a plan that resolves the issue.

26587 Query Execution, Query Optimizer 4.3.11.0 Pivotal Query Optimizer (PQO) caused a Greenplum Database PANIC for queries that contain a scalar subquery with a single constant table child and the scalar subquery returns an outer reference. For those queries, PQO incorrectly processed the subquery.

This issue has been resolved.

26524 Scripts: Backup and Restore 4.3.11.0 In some cases, when restoring database data, the Greenplum Database gpdbrestore utility incorrectly reported errors when restoring indexes, constraints, and other metadata when restoring the data. The restore operation completed successfully.

Now, the incorrect errors are no longer reported.

26504 Scripts: gptransfer 4.3.11.0 In some cases, a transfer operation failed when the gptransfer utility did not correctly compare source and destination partitioned tables. The utility returned a message stating that the column layout of the two tables was different. The failure occurred if a column was dropped and then added back to the source partitioned table and then the DDL for the table was used to create the destination table.

Now, the gptransfer utility correctly compares the column layout of the partitioned tables.

26469 Storage: Vacuum/ Reindex/ Truncates 4.3.11.0 Executing gpstop -i after reindexing the pg_class catalog table in the template1 database could leave the database in an inconsistent state.

This issue has been resolved.

26187 Query Execution 4.3.11.0 In some cases, long running sessions were consuming memory on Greenplum Database and the sessions were idle on the master segment and did not have connections to segment instances.

Memory management has been improved to better handle long running transactions.

134801867 Transaction Management 4.3.11.0 In some cases, Greenplum Database did not handle the processing of two phase commits properly when the first phase of the two phase commit failed. This issue caused the pg_xlogs to grow to a large size.

The processing of two phase commits has been enhanced to handle the failure properly.

133827909 Query Optimizer 4.3.11.0 Pivotal Query Optimizer (PQO) returned incorrect results for queries that contained a predicate on a partitioned table partition key column and a subquery that contained a LIMIT clause. PQO incorrectly pushed the predicate into a subquery.

This issue has been resolved.

132718373 Query Optimizer 4.3.11.0 When Pivotal Query Optimizer (PQO) and the server configuration parameter optimizer_parallel_union are both enabled, some queries that contained a UNION operator would incorrectly fall back to the legacy query optimizer.

Now, PQO executes the queries.

132714585 Management Scripts 4.3.11.0 In some cases, after performing an incremental segment recovery (running the Greenplum Database utility gprecoverseg without the -F option), did not recover the Greenplum Database segment correctly and caused a SIGSEGV.

Now, gprecoverseg correctly performs incremental segment recovery.

131931989 S3 External Table 4.3.11.0 When an Greenplum Database readable external table is defined with the s3 protocol and the table definition uses the HEADER option, Greenplum Database did not correctly process table header information when a Greenplum Database segment read data from multiple files.

This issue has been resolved.

131931971 S3 External Table 4.3.11.0 When an Greenplum Database readable external table is defined with the s3 protocol, Greenplum Database did not correctly read multiple files when the last line of a file did not contain a trailing return character. The last line of the first file (without a trailing return character) was joined to first line of the second file and was read as a single line.

Now, files without a trailing return character are processed correctly.

131309733 Query Execution 4.3.11.0

The EXPLAIN ANALYZE command did not report the correct number of segments that spill workfiles for a query, if the query contains an ORDER BY clause and the query's sort operator spilled workfiles, and the server configuration parameter gp_enable_mk_sort is off.

Now, EXPLAIN ANALYZE reports the correct number of segments that execute sort operations and spill workfiles.

130407893 Transaction Management 4.3.11.0 If a Greenplum Database segment failed during two phase transaction processing, the transaction remained in a uncompleted state and was cleaned up only during a Greenplum Database restart. In many cases, this caused high disk consumption by the Greenplum Database xlog process.

This issue has been resolved.

130121327 Query Optimizer 4.3.11.0 When Pivotal Query Optimizer (PQO) and the server configuration parameter optimizer_parallel_union are both enabled, PQO evaluated query plans that use only parallel union all and did not evaluate plans that use serial union all.

Now, both types of plans are now evaluated.

129871531 Query Optimizer 4.3.11.0 When Pivotal Query Optimizer (PQO) is enabled, PQO did not infer additional predicates based on the join and select predicates for some queries that contain a LIMIT clause over subqueries and a join. This resulted in the generated query plan being sub-optimal.

This issue has been resolved.

4.3.10.0 Release

Table 23. Resolved Issues in 4.3.10.0
Issue Number Category Resolved In Description
26607 Query Optimizer 4.3.10.0 When the SELECT clause of a CTAS (CREATE TABLE AS SELECT) command contains row_number window function, Pivotal Query Optimizer (PQO) generated a query plan that returned wrong results. PQO incorrectly generated a broadcast motion in the query plan which caused the incorrect duplication of row_number.

This issue has been resolved.

26574 Query Execution 4.3.10.0 Pivotal Greenplum Database would throw an error message when it tried to close a temporary file (for example, a bfz workfile) that did not exist. If the number of error messages exceeded the ERRORDATA_STACK_SIZE setting, then Greenplum Database triggered a PANIC signal. The issue has been resolved by throwing warnings instead of errors for missing temporary files.
26570 Query Planner 4.3.10.0 With the fix to Issue 26163 introduced in Greenplum Database 4.3.8.0, the legacy query optimizer now uses default statistics when they are missing for a table. Because this can lead to sub-optimal plans, a new configuration parameter, gp_enable_relsize_collection, has been introduced. Enable this parameter to use the estimated size of a table (pg_relation_size function) if there are no statistics for the table. Note that enabling this parameter will increase query optimization time and resource queue usage in heavy workloads.
26557 Query Execution 4.3.10.0 When GPORCA was enabled for a query having AND/OR predicates on a table with both B-tree and bitmap indexes, then the query executor could hit a SEGV caused by double-freeing the memory that was used by the hash bitmap. This issue has been resolved.
26553 Dispatch 4.3.10.0 A memory freeing problem during gang cleanup could leave active query executors (QEs) behind after a portal gang was destroyed. This could cause COPY operations to fail with the error:
"PANIC","XX000","Unexpected internal error: Master process received signal SIGSEGV"
The issue has been resolved by waiting for all QEs to complete and removing the gang destroy action from the gang cleanup function.
26537 Management Scripts 4.3.10.0 The gptransfer utility failed when transferring any table with a column name containing capital letters or quotes if that column was part of the distribution key. The failure occurred when gptransfer attempted to create an external web table for the database table because the DISTRIBUTED BY clause was not properly escaped.

This issue has been resolved.

26544 Storage: Vacuum 4.3.10.0 Performing a VACUUM FULL operation could cause a PANIC condition if certain catalog tables referenced unused pages. This issue has been resolved by skipping the VACUUM operation on catalog tables that might contain unused pages. Because only catalog tables are skipped, this change does not affect the overall storage that is reclaimed in a VACUUM operation.
26503 Management Scripts: Backup and Restore 4.3.10.0 When performing a backup operation with the gpdbrestore utility to a Veritas NetBackup server, the operation failed if the gpdbrestore utility option --netbackup-keyword was specified and the length of the value was longer than the maximum.

This issue has been resolved. Now the utility returns an error. See Product Enhancements.

26494 Management Scripts 4.3.10.0 The gpdbrestore utility -S option specifies a database schema to restore from a backup. The utility restored tables belonging to the schema, but not other supported objects, such as views, sequences, user-defined functions, and indexes.

The utility now restores those object qualified by the specified schema. See Product Enhancements.

26484 Query Planner 4.3.10.0 The legacy query planner generated wrong results for correlated queries having predicates that were not of the form A op B, where A and B are expressions. This issue has been resolved
26477 Query Execution 4.3.10.0 For some queries that spilled to disk, Greenplum Database would leak a counter that measures the bytes written to disk, if the query was aborted mid-execution because of an error. This issue has been resolved by accounting for this scenario and updating the counter correctly.
26461 gpfdist 4.3.10.0 The maximum value for the gpfdist -t and -w options has been increased to 7200 seconds (2 hours) to allow for long-lived workflow connections.
26450 Query Optimizer 4.3.10.0 When GPORCA (PQO) was enabled, Greenplum Database generated wrong results in the rare case where a query had an outer join with the inner side of the join as a partitioned table, and an IsNull join condition on the partition column. This issue has been resolved.
132643231 gphdfs 4.3.10.0 Specifying an external table with the gphdfs protocol and with these characters \, ', <, > was a possible security vulnerability.

This issue has been resolved. Now the characters are not allowed with the gphdf protocol. See Changed Features.

130382403 Management Scripts 4.3.10.0 When the gprecoverseg utility was run with the -r option and some of the Greenplum Database segment instances were in resynchronization mode, the utility would hang indefinitely.

The issue has been resolved.

129977107 Query Optimizer 4.3.10.0

In Pivotal Greenplum Database, there is a preprocessing step inside the Pivotal Query Optimizer (PQO) that prunes unused computed columns. In scenarios where the query is deeply nested (for example, in a common table expression, in nested views, or in nested subqueries) this stage could suffer from unnecessarily copying internal metadata. This delayed the plan generation time.

This issue has been resolved by improving the preprocessing step. The query optimization time has been significantly improved for deeply-nested queries.

128520573 Query Optimizer 4.3.10.0 For queries that have correlated subqueries where the outer query contains a filter on a partition table, GPORCA (PQO) would fallback to the legacy query optimizer. This issue has been resolved, and the PQO now successfully generates the query plan.
127773869 Catalog & Storage 4.3.10.0 Executing gpstop -i after reindexing the pg_class catalog table could leave the database in an inconsistent state. This issue has been resolved.
127488131 Catalog & Storage 4.3.10.0 Improved the error message that is displayed when a waiting TRUNCATE statement is canceled.
127333267 Query Optimizer 4.3.10.0 When GPORCA (PQO) is enabled, Greenplum Database generated wrong results in the rare circumstances where a query satisfied all of these criteria:
  • The query contained a projection defined as a scalar subquery with a LIMIT of 1 or 0.
  • The subquery contained a count aggregate over a subexpression.
  • The subexpression returned an empty set.

This issue has been resolved.

121272245 Catalog & Storage 4.3.10.0 Certain database system files, such as pg_xlog and pg_clog, were always copied from a primary to mirrors during start up. The code has been modified to copy these files only when necessary.

4.3.9.x Releases

Table 24. Resolved Issues in 4.3.9.x
Issue Number Category Resolved In Description
26059 Management Scripts: master mirroring 4.3.9.1 gpinitstandby could take a long time to complete if large files were stored under gpperfmon/logs, because gpinitstandby unnecessarily synchronized those files to the standby host. Also, gpinitstandby excluded the gp_dumps directory from synchronization, whereas the correct directory name to exclude is db_dumps.

This issue has been resolved by excluding the gpperfmon/logs and db_dumps directories from gpinitstandby synchronization.

26348, 26463 Locking, Signals, Processes 4.3.9.1 When SSL connections were enabled, a query could hang in a deadlock situation wherein the query could not be terminated using the pg_terminate_backend function. The associated idle session would remain in pg_lock.

This issue has been resolved.

26405 Locking, Signals, Processes 4.3.9.1 Poor spinlock performance in the Master segment could result in high CPU utilization (up to 100%) under highly-concurrent workloads.

This issue was resolved by patching the associated lock implementation.

26456 gpperfmon 4.3.9.1 Certain commands and operations were added to the gpperfmon database query_now and query_history tables with anomalous tsubmit dates in 1969. These commands and operations include EXPLAIN/ANALYZE and COPY commands and, depending on the logging level, some service and function calls. This has been fixed.
26481 Storage: Segment Mirroring 4.3.9.1 In rare cases, inserts into an append-only table could fail and then cause queries against the table to fail with the error:
ERROR: Header checksum does not match 

This issue has been resolved.

26506 Locking, Signals, Processes 4.3.9.1 Executing an INSERT command to insert data to an external table as part of a subtransaction could cause GPDB to attempt to free memory that was already freed. This would lead to a PANIC in the associated segment, followed by a PANIC in the master segment.

This issue has been resolved.

26507 Dispatch 4.3.9.1 A memory overflow problem could cause a corrupted double-linked list in glibc logging, and ultimately result in the master segment crashing with a signal 6: Aborted error.

This issue has been resolved.

113452283 Persistent Table Rebuild 4.3.9.1 Greenplum Database could fail to rebuild persistent tables if a mirror was down and:
  • the associated mirror segment directory was empty (for example, due to a failed gprecoverseg -Fa command), or
  • a file in the the mirror directory was corrupted or deleted.
If either of the above conditions occurred when a mirror was down, then the rebuild operation would fail with a message similar to:
[CRITICAL]:-gppersistent_backup.py failed. 
(Reason='Missing per-database persistent files from source directory.') exiting...
This issue has been resolved. The persistent table rebuild code now recovers the primary segment, even if a mirror is unavailable and required files on the mirror are missing or corrupt.
26513 Query Optimizer 4.3.9.0 When Pivotal Query Optimizer (PQO) is enabled, PQO returned wrong results for a query if the query contains computed columns and also contains a set returning function. Wrong results were returned because PQO incorrectly collapsed the computed columns into a single compute node.

This issue has been resolved.

26474 Dispatch, Languages: R PLR 4.3.9.0 Some queries that execute a PL/R function generated a PANIC on a Greenplum Database segment. The PANIC was caused when PL/R incorrectly handled of data that is not of R datatype CHARSXP.

Handling of data that is not of type CHARSXP been improved. See changed Changed Features.

26465 Scripts: Backup and Restore 4.3.9.0

When a backup operation was performed with the Greenplum Database utility gpcrondump with the -S option to exclude a schema from the backup, and a temporary schema (with the name pg_temp_nnn) existed in the database, the utility incorrectly attempted to back up the temporary schema and corresponding tables. This caused the utility to hang during the backup operation.

This issue has been resolved. The utility does not back up temporary schemas and their corresponding tables.

26454 Query Execution, Resource Management 4.3.9.0 In some cases during Greenplum Database query execution, data overflow could have occurred when a tuple had more than 4096 null attributes. This would cause a SIGSEGV.

This issue has been resolved. To prevent data overflow, the data types that are used in the calculations have been changed.

26453 Query Execution 4.3.9.0 In some cases, a Greenplum Database error occurred during the execution of a query plan that was generated by Pivotal Query Optimizer (PQO). The error was caused by a resource that was incorrectly locked more than once. The incorrect locking was caused by PQO transformation rules and the query plans that were produced by PQO.

The incorrect locking issue has been resolved.

26452 Query Optimizer 4.3.9.0 When enabled, GPORCA (PQO) caused a SIGSEGV for queries where both features are present.
  • The query contains a predicate that performs a greater-than or less-than comparison between a table column and a scalar subquery
  • An index is defined on the table column.

Now, for queries those properties, PQO falls back to the legacy planner.

26451 Resource Management 4.3.9.0 In some cases, Greenplum Database locks that were used by the Greenplum Database resource queues were not handled properly. This causes a SIGSEGV.

Greenplum Database handling and monitoring of locks that are used by resource queues has been improved.

26439 Storage: Transaction Management 4.3.9.0

In some cases, declared foreign key constraints on append-optimized (AO) tables were triggered when the legacy planner executed DML statements on those AO tables. This resulted in errors when executing DELETE and UPDATE commands.

This issue has been resolved. The legacy planner no longer triggers declared foreign key constraints on AO tables during DML execution.

26422 Query Execution 4.3.9.0 In some cases, when a partitioned table that is defined with a default partition was altered with the ALTER TABLE command that contains the SPLIT DEFAULT PARTITION clause, Greenplum Database generated a PANIC.

This issue has been resolved.

26421

26467

Resource Management 4.3.9.0 In some cases, when a query was cancelled and the query was managed by a Greenplum Database resource queue, Greenplum Database generated a PANIC from a segment process crash and caused a postmaster reset.

This issue has been resolved.

26399 Query Execution 4.3.9.0 Greenplum Database generated a PANIC during query execution under these conditions:
  • The query contains a SharedInputScan that has a Sort child node.
  • The Sort spills to disk.
  • The query execution process receives a request from the master to finish the execution process when query execution is complete.

The PANIC was caused by improper handling of the request to finish the query execution.

This issue has been resolved. Now, the request to finish the query execution is handled properly.

26397 Query Execution 4.3.9.0

Greenplum Database did not process the query correctly if a CASE statement in the form of CASE field WHEN IS NOT DISTINCT FROM results appears in the GROUP BY clause of a query.

This issue might also arise if the DECODE function is in the GROUP BY clause because, in some cases, Greenplum Database transforms the DECODE function into a CASE statement of the specified form during query processing.

This issue has been resolved.

26394 Dispatch 4.3.9.0 In some situations, Greenplum Database generated log messages stating that it was a rejecting TCP connection to master. The log message was generated when a Greenplum Database query dispatcher process on the master attempted to connect to the postmaster process on the master with TCP connection.

This issue has been resolved. Now, Greenplum Database uses a UNIX domain socket for the connections.

26385 Catalog and Metadata, Query Execution 4.3.9.0 The function convert(string, [src_encoding,] dest_encoding) returned an error message or generated a Greenplum Database PANIC when the function was passed only the string and dest_encoding arguments (the optional src_encoding argument was not passed).

This issue has been resolved. Now, the function performs correctly.

26376 Query Execution, Storage: Transaction Management 4.3.9.0 In some cases, when compression is enabled for Greenplum Database workfiles (the server configuration parameter gp_workfile_compress_algorithm is set to zlib), Greenplum Database generated an ERROR on a Greenplum Database segment that stated a memory block containing chunk chunk-ID could not be found. The error was caused by improper memory handling during query execution.

This issue has been resolved.

26365 Query Optimizer 4.3.9.0 For some queries that contained a ROW expression with data type of varchar with a length of 7 larger, Greenplum Database generated a SIGSEGV error. The error was caused by improper initialization of record types during query execution.

This issue has been resolved.

26351 Query Optimizer 4.3.9.0 When a query contained a subquery with a set-returning function, Pivotal Query Optimizer (PQO) generated a query plan that returned wrong results. PQO incorrectly pulled the function up to a higher location in the query plan when generating the query plan.

This issue has been resolved.

26325 Query Execution 4.3.9.0 When the Greenplum Database server configuration parameter lc_numeric was set in a psql session to control the formatting of displayed numeric data, Greenplum Database ignored the value of the parameter after a short time and did not format the numeric data correctly.

This issue has been resolved.

26319 Global Persistent Objects 4.3.9.0 In some cases, the free tuple ID (TID) list that is used by Greenplum Database became corrupted and returned a message stating that the number of free TIDs did not match the TID order number.

This issue has been resolved. The management of the free TID list has been improved.

26315 Query Execution 4.3.9.0 In some cases with Pivotal Query Optimizer enabled, Greenplum Database generated a PANIC during query plan generation that stated the ERRORDATA_STACK_SIZE was exceeded. The PANIC was caused by improper error handling during query plan generation.

This issue has been resolved.

26314 Query Optimizer 4.3.9.0 When Pivotal Query Optimizer (PQO) is enabled and after a column with a NOT NULL constraint was dropped from a table, an INSERT command for the table generated a query plan that incorrectly enforced a column NOT NULL constraint.

This issue has been resolved.

26305 Dispatch 4.3.9.0 In some situations when performing a VACUUM ANALYZE operation, Greenplum Database generated PANIC. This was caused by improper memory handling during the operation.

This issue has been resolved.

26289 Query Optimizer 4.3.9.0 In certain cases, Pivotal Query Optimizer (PQO) generated a query plan that contained a gather motion on top a plan that was replicated using a broadcast motion. During query execution, the plan generated a SEGV error.

In this release, PQO generates the correct plan and does not fall back to the legacy planner for these cases.

26256 Scripts: Backup and Restore 4.3.9.0 When backing up a partitioned table, the Greenplum Database gp_dump utility also incorrectly backed up a table with the same name that was in a different schema.

This issue has been resolved.

26246 Dispatch 4.3.9.0 In Greenplum Database, the maximum number of columns that can be specified in the DISTRIBUTED BY clause of the CREATE TABLE command is 6000. Greenplum Database did not properly handle a CREATE TABLE command that specified more than 6000 columns in the DISTRIBUTED BY clause.

This issue has been resolved.

25914 Scripts: Backup and Restore 4.3.9.0 If the database owner ID contained special characters, a restore operation failed when using Greenplum Database utility gpdbrestore.

This issue has been resolved.

25565 Scripts: recoverseg 4.3.9.0 During an incremental segment recovery, if the Greenplum Database gprecoverseg utility detected a corrupted change tracking log that was used for the recovery, the utility did not inform the user that the log was corrupted and that a full recovery was required.

This issue has been resolved. Now, if the utility detects a corrupt log during incremental recovery, the message states the error and that a full recovery is required.

GPSQL-3272 Query Optimizer 4.3.9.0 When optimizing queries that contain WITH clauses (CTE queries) and predicates involving volatile functions, Pivotal Query Optimizer (PQO) incorrectly pushed the predicates into the CTE producer. In these cases, incorrect results were returned.

This issue has been resolved. PQO no longer pushes predicates with volatile functions into a CTE producer.

120748249 S3 External Table 4.3.9.0 Reading data from a Greenplum Database external table that specified the s3 protocol failed if the S3 bucket in the LOCATION clause did not contain a prefix.

This issue has been resolved. The S3 bucket location does not require a prefix.

120726595 S3 External Table 4.3.9.0 If a readable external table was created with the s3 protocol, reading data from a gzip format file that is larger than the chunksize value specified in the s3 configuration file returned incorrect data.

This issue has been resolved.

119975145 S3 External Table 4.3.9.0 If a readable external table was created with the s3 protocol, and the protocol could not allocate sufficient memory during a read operation, an error message might not have been returned. An incorrect empty result set might have been returned.

This issue has been resolved.

119939221 Query Execution 4.3.9.0 When a Greenplum Database query plan contained an InitPlan that contained a SET command, Greenplum Database improperly deleted a query execution process during query execution. This caused Greenplum Database to generate a SIGSEGV.

This issue has been resolved.

116339321 Dispatch 4.3.9.0 In some cases, Greenplum Database dispatcher leaked a small amount of memory when cleaning up distributed worker processes after query execution completed.

This issue has been resolved.

4.3.8.x Releases

Table 25. Resolved Issues in 4.3.8.x
Issue Number Category Resolved In Description
26281 Query Execution 4.3.8.1 Some queries that contain a window operator that creates a large number of window partitions might have consumed a significant amount of memory due to a memory leak in the per-partition window state.

This issue has been resolved.

26278 Storage: Transaction Management 4.3.8.1 In some cases, inserting data into an append-optimized columnar table with RLE compression returned an ERROR stating that the physical datum size would exceed maximum data block size.

This issue has been resolved.

26257 Storage: Filespace / Tablespace 4.3.8.1 In some situations, the server configuration parameter gp_default_storage_options could be set so that a table that was created could have storage options on the Greenplum Database master that were different than the options on the Greenplum Database segments.

This issue has been resolved.

26236 Dispatch 4.3.8.1 In some cases, a Greenplum Database master reset occurred when Greenplum Database did not handle an out of memory condition correctly.

This issue has been resolved. Greenplum Database now handles this out of memory condition correctly.

26226 Query Optimizer 4.3.8.1 For queries that contain a NOT operation over a quantified subquery where the scalar comparison operator did not have an associated inverse function (for example, inet operator <<), the Pivotal Query Optimizer (PQO) generated a SIGSEGV error.

This issue has been resolved. PQO generates the correct plan without an error.

26217 Interconnect 4.3.8.1 In some cases while performing a Greenplum Database upgrade from 4.2 release to a 4.3 release, the upgrade process failed and generated a SIGSEGV error. The error was caused because the upgrade process did not handle interconnect thread detection correctly.
26212 Resource Management 4.3.8.1 Greenplum Database displayed an warning message about a tuple descriptor leak when loading data into some partitioned tables if the table was altered and the ALTER TABLE command specified one of the following clauses: DROP COLUMN, ADD COLUMN, EXCHANGE PARTITION, SPLIT PARTITION.

This issue has been resolved.

26205 Storage: Segment Mirroring 4.3.8.1 In some cases, Greenplum Database became unresponsive and returned a PANIC after a timeout period while waiting to write to an in-memory file replication log.

This issue has been resolved. File replication logs are written to the directory pg_log of the Greenplum Database master and each segment instance data directory.

26162 Query Execution 4.3.8.1 In some cases, an out of memory error occurred when compressed workfiles were used during Greenplum Database query execution.

Greenplum Database has enhanced the tracking of memory that is used when query execution uses compressed workfiles. See Performance and Stability Enhancements.

26112 Locking, Signals, Processes 4.3.8.1 In some situations, a query that was running only on the Greenplum Database master host could not be terminated with the pg_cancel_backend.

This issue has been resolved.

25380 Backup and Restore 4.3.8.1 When the Greenplum Database utility gpdbrestore performed a restore operation and the --redirect option was specified, the utility failed if the name of the database that was originally backed up was changed.

This issue has been resolved. The restore operation does not require the existence of the original database.

23394 Query Execution 4.3.8.1 In some cases when a query plan contains two SharedScan operators that were consuming results from a Materialize operator, and the SharedScan operators are executed in different slices, a deadlock might have caused the query to hang indefinitely.

This issue has been resolved.

GPSQL-3121 Query Execution 4.3.8.1 In some cases, splitting a partition of a partitioned table with the ALTER TABLE SPLIT PARTITION command returned a warning message stating that the system leaked a tuple descriptor reference.

This issue has been resolved.

26228 Query Optimizer, Query Planner 4.3.8.0 Some queries generated a segmentation fault if the query was executed by the legacy query planner and the query performed a correlated execution against external web tables.
Now, the queries returns this message that instruct the user to use GPORCA.
Rescan is not supported for web external table: table_name.
Please use 'set optimizer=on' as a work around
and 'set optimizer_enable_master_only=on' if accessing catalog tables. 

For information about the server configuration parameter optimizer_enable_master_only_queries. See New and Changed Parameters.

For information about the server configuration parameters, see the Greenplum Database Reference Guide. For information about GPORCA, see the Greenplum Database Administrator Guide.

26224 Catalog and Metadata, Storage: Transaction Management 4.3.8.0 For DDL operations, additional validations for index operations on system catalog tables have been introduced to prevent improper catalog modifications in the case of corrupted indices on catalog tables.
26221 Interconnect 4.3.8.0 For some queries, a Greenplum Database query dispatcher received an unexpected message from a Greenplum Database query executor. This caused Greenplum Database to log an incorrect log message that contains this text.
Motion layer node 1 cleanup

This issue has been resolved.

26201 Management Scripts: gptransfer 4.3.8.0 In some cases the Greenplum Database utility gptransfer utility failed when the --partition-transfer option was specified. The error message did not clearly indicate the error.

Now, the utility returns an error message that describes the issue.

26198 Dispatch 4.3.8.0 In some cases, Greenplum Database did not properly handle an out of memory condition that occurred on the master. This caused a segmentation fault.

This issue has been resolved. Greenplum Database reports an ERROR for these out of memory situations.

26176 Query Optimizer 4.3.8.0 For some queries that contained a computed column and the column contained a set returning function, Pivotal Query Optimizer incorrectly considered the column as an unused column and dropped the column during query optimization.

This issue has been resolved.

26163 Query Execution 4.3.8.0 When querying tables without statistics, the legacy query optimizer triggered on demand statistics computation. This increased the query optimization time and could exhaust resource queues under heavy workload.

This issue has been resolved. The legacy query optimizer now uses default statistics when they are missing.

If the generated plan is sub-optimal, you can run ANALYZE to generate current statistics for the table.

26158 Dispatch 4.3.8.0 For some queries, a PANIC was generated on the Greenplum Database master due to the improper handling of strings when an internal connection was started.
26087 Dispatch 4.3.8.0 A Greenplum Database PANIC was generated by a query that executed a function that was run on multiple segments and the function attempted to set a Greenplum Database server configuration parameter with the SET command.
26078 Management Scripts: gptoolkit 4.3.8.0 In the Greenplum Database gp_toolkit administrative schema, the rsqwaiters column of the gp_resqueue_status table returned wrong value.
26052 DDL and Utility Statements 4.3.8.0 In some situations, a running Greenplum Database query could not be terminated with the functions pg_cancel_backend or pg_terminate_backend.

This issue has been resolved. The functions will cancel a running query.

26045 Backup and Restore 4.3.8.0 When performing a backup operation with the Greenplum Database gpcrondump utility that specified a file with the --schema-file option, the utility deleted the file after completing the backup.
25924 Query Execution 4.3.8.0 When Pivotal Query Optimizer (PQO) is enabled, a plan that generated bitmap index scans caused a query execution crash. This issue has been resolved and plans that contain bitmap index scan perform as expected with PQO.
Note: In previous releases, generating plans with bitmap index scans is disabled for Pivotal Query Optimizer.
25829 Dispatch 4.3.8.0 If the name of a Greenplum Database contains a quote character, SQL commands that were run in the database returned an error if the commands accessed Greenplum Database segments.
25821 Backup and Restore 4.3.8.0 Greenplum Database gpcrondump utility did not back up tables with names that are all uppercase characters.
25147 Query Optimizer 4.3.8.0 When GPORCA executed an ALTER TABLE command with REORGANIZE clause, if the distribution policy of the table is being changed to random distribution (with the DISTRIBUTED RANDOMLY clause), the generated plan did not force a reorganization of the tuples across Greenplum Database segments.

This issue has been resolved. Greenplum Database falls back to the legacy query optimizer to execute the command.

4.3.7.x Releases

Table 26. Resolved Issues in 4.3.7.x
Issue Number Category Resolved In Description
26220 Query Planner 4.3.7.3 In some cases, the Greenplum Database legacy planner did not properly the manage common table expressions (CTEs) that were in queries that contain nested common table expressions. This caused Greenplum Database to generate a PANIC.
26127 Query Execution 4.3.7.3 Greenplum Database did not properly process some queries that contained window functions. This caused Greenplum Database segmentation faults on the master and segments.
26182 Query Execution 4.3.7.2 In some cases, executing a query that contained nested functions caused a Greenplum Database segment reset.
26178 Storage: Segment Mirroring 4.3.7.2 In previous Greenplum Database releases, this message is defined as a LOG level message.
FTS: found bad segment with dbid ID

Now, the message is defined as a WARNING level message.

26107

26115

Query Planner 4.3.7.2 For some queries that returned a large result set that required sorting, Greenplum Database generated a segmentation fault because the query dispatcher and query executor did not properly handle the sorting of data within Greenplum Database.

This issue has been resolved.

26093 Query Planner 4.3.7.2 For some queries that contain common table expressions (CTEs), the legacy planner caused Greenplum Database to generate a PANIC when the unsupported server configuration parameter gp_cte_sharing is set to on. When the parameter is enabled, there are known issues with queries that contain CTEs. The default value of the parameter is off.

The server configuration parameter guc gp_cte_sharing controls the ability of the legacy planner to share a CTE during query execution (execute a CTE once and use it multiple times). When the parameter is set to off, CTE sharing is disabled.

26086 Query Optimizer 4.3.7.2 In some cases when GPORCA executed a query against a table from which a column was dropped, Greenplum Database returned a message that incorrectly stated that some columns did not have statistics and recommended that the table should be analyzed.

This issue has been resolved. The message is no longer returned.

26135 Backup and Restore 4.3.7.1 In some cases, Greenplum Database utility gpcrondump performance was poor because the utility was performing checks for database metadata, such as indexes and triggers, on both the Greenplum Database segments and the master.

Performance has been improved by not performing the checks on the segments.

26141 Management Scripts: gpssh 4.3.7.0 In some cases, the Greenplum Database utility gpssh did not wait long enough for a connection to a Greenplum Database host and returned an error when attempting to connect to a host.

This issue has been resolved. To minimize connection failures, the amount of time that gpssh waits for a connection has been increased.

26123 DML 4.3.7.0 In some situations, an SQL query that required data from multiple segments did not return results when executed. The same query returned results when executed at a later time.

This issue has been resolved. During query execution, the communication between Greenplum Database segments has been improved.

26122 Query Optimizer 4.3.7.0 When GPORCA (PQO) is enabled (optimizer=on), PQO did not distribute table data for ALTER TABLE commands that contained a distribution clause.

The issue has been resolved. The PQO falls back to the legacy planner for ALTER TABLE commands that contain a distribution clause to ensure table data is rebalanced across the Greenplum Database segments.

26116 GPHDFS 4.3.7.0 The gphdfs protocol returned an exception when writing a NULL value to a Parquet format file.

This issue has been resolved. The gphdfs protocol writes NULL values to a Parquet file without errors.

26104 Storage: Access Methods 4.3.7.0 Greenplum Database memory increased while inserting a large amount data into an append-optimized table because Greenplum Database did not properly release memory.

Greenplum Database memory management has been improved when inserting data into append-optimized tables.

26091 Query Execution 4.3.7.0 In some cases, the mregr_* functions returned incorrect results due to incorrect initialization.

This issue has been resolved. Now, newly created matrix values are reset to zero before performing the calculation.

26076 Query Optimizer 4.3.7.0 When GPORCA (PQO) is enabled (optimizer=on), Greenplum Database generated a PANIC for some queries that contain the FULL OUTER JOIN clause on partitioned tables and predicates on partitioning columns.

This issue has been resolved. The PQO falls back to the legacy planner for the specified queries.

26075 Loaders: gpload 4.3.7.0 The Greenplum Database Loaders for Red Hat Enterprise 5 and 6 did not contain the library libyaml-0.so.1.

This issue has been resolved.

26063 Query Optimizer 4.3.7.0 GPORCA performance was poor for queries that contain distinct qualified aggregates (DQA) without a grouping column when the table is not distributed on the columns used by the DQA.

Performance has been enhanced, for queries that contain DQA.

26050 GPHDFS 4.3.7.0 A Greenplum Database PANIC occurred when reading data from an Avro file that contains Unicode characters with the gphdfs protocol.
26028 Query Planner 4.3.7.0 In some situations, a query that was running only on the Greenplum Database master host could not be terminated with the pg_cancel_backend or pg_terminate_backend functions.
26023 Backup and Restore 4.3.7.0 In some cases, the Greenplum Database gpdbrestore utility incorrectly restored tables in the default schema instead of their expected schema. The incorrect restore was caused by the Greenplum Database gpcrondump utility incorrectly backing up the database schema information.

This issue has been resolved.

26016 Storage: Transaction Management 4.3.7.0 Greenplum Database dropped all user connections due to a Greenplum Database PANIC when a user with a default tablespace defined attempted to log in to Greenplum Database and an error occurred during Greenplum Database tablespace assignment for the user.
26014 Backup and Restore 4.3.7.0 The Greenplum Database utility gpdbrestore returned an error and did not restore a table from a backup if the --truncate option was specified and the table to be restored did not exist in the target database.

This issue has been resolved. The utility restores the table from the backup and returns a warning message.

26007

25972

Query Execution 4.3.7.0 For some SQL commands that contain grouping operations such as CUBE and ROLLUP and GROUPING SETS, Greenplum Database did not handle memory properly when executing the command and caused a Greenplum Database segmentation fault.
25988 Backup and Restore 4.3.7.0 In some cases, the Greenplum Database utility gpdbrestore did not return a warning or error when a restore operation did not successfully insert data.

This issue has been resolved.

25904 Backup and Restore 4.3.7.0 The Greenplum Database utility pg_dumpall could not dump filespaces. The -f option was used for both providing an output file name and for dumping filespaces, and the option could only be used to provide an output file name.

This issue has been resolved. Now the utility uses the -F option for dumping filespaces. The -f option is used to provide an output file name.

25844 Query Planner 4.3.7.0 When a Greenplum Database system is running multiple segments, the Greenplum Database legacy planner estimate for the number of rows for the table pg_partition_rule did not account for multiple segments.
25584 Query Execution 4.3.7.0 In some situations, a running Greenplum Database query could not be terminated with the functions pg_cancel_backend or pg_terminate_backend.

Now, pg_terminate_backend will cancel a running query.

25385 Client Access Methods and Tools 4.3.7.0 In some situations when Greenplum Database hung when sending data to a client application, the connection to the application could not be terminated with the pg_terminate_backend function. The function has been improved.
25264 Backup and Restore 4.3.7.0 The Greenplum Database utility gpdbrestore performance was poor when append-optimized tables were restored. The utility updated tuple count information for all append-optimized tables in the database, not just the tables that were restored.

The performance has been improved. The utility updates only append-optimized tables that are restored.

4.3.6.x Releases

Table 27. Resolved Issues in 4.3.6.x
Issue Number Category Resolved In Description
26053 Query Optimizer 4.3.6.2 A Greenplum Database PANIC occurred when GPORCA attempted to optimize a query that contained derived tables that have N-ary join operations with no join predicates. The PANIC occurred because statistics were not derived for a small subset of partial plans that were generated.

This issue has been resolved.

26021 Query Execution 4.3.6.2 For a query that contains the sequence operator nextval, the Greenplum Database query execution plan executes the nextval function either on segment nodes or on the master node. If the sequence definition allows caching sequence values on segment instances,the nextval function might have incorrectly returned the maximum or minimum value when the sequence values were exhausted.

Now, the nextval function correctly returns the error nextval: reached maximum/minimum value."

25983 Upgrade/Downgrade 4.3.6.2 If Greenplum Database 4.2.x was not shut down properly before performing an upgrade with the Greenplum Database 4.3 gpmigrator_mirror utility, the utility would incorrectly detect the system was up and would fail. Also, when the Greenplum Database gpmigrator_mirror utility did not successfully upgrade a Greenplum Database system from, the rollback process failed when the original data directory contained a symbolic link to a directory.

These issues have been resolved.

25977 Query Optimizer 4.3.6.2 For some queries that contain multiple renamed columns of a single column, and the renamed columns are used either in a common table expression (a WITH clause) or in a UNION or UNION ALL operation, Pivotal Query Optimizer generated an incorrect plan that filters out null values.

This issue has been resolved.

25960 Upgrade/Downgrade 4.3.6.2 If the Greenplum Database 4.3 gpmigrator_mirror utility detected a previously failed upgrade process when upgrading a Greenplum Database 4.2.x system, the utility restarted the upgrade process from the point of the previous failure.

Now, the utility displays a prompt and lets the user choose to either continue with the previous upgrade or start a new upgrade.

Note: A new upgrade should be started only if Greenplum Database is in a stable, upgradable state. For example, the system has been successfully recovered from a previous upgrade that failed.
25959 DDL and Utility Statements 4.3.6.2 Enhanced the Greenplum Database DROP operation for tables. The enhancement prevents some catalog inconsistencies between the Greenplum Database master and segment instances that could have occurred when tables were dropped.
25949 Management Scripts: analyzedb 4.3.6.2 The Greenplum Database analyzedb utility returned an error if the utility encountered a table or schema name that contained uppercase characters (the name was enclosed in quotes and contained upper case characters).

This issue has been resolved.

25943 Backup and Restore 4.3.6.2 In some cases, the Greenplum Database gpcrondump utility hangs and holds a lock on the pg_class catalog table when the utility encounters a failure on a segment host.
25897 Management Scripts: gptransfer 4.3.6.2 In addition to existing data validation, the Greenplum Database gptransfer utility compares the table row counts between destination and source tables. The new option --no-final-count disables the validation.

For information about the new gptransfer validation, see Product Enhancement.

25875 Catalog and Metadata 4.3.6.2 In some cases, when Greenplum Database performed an ALTER TABLE RENAME and an ANALYZE operation on a table concurrently with a DROP TABLE operation on the same table, the table was dropped only on the Greenplum Database master. The table was not dropped on Greenplum Database segments.
25837 Storage: Access Methods 4.3.6.2 In some cases, inserting data into an append-optimized columnar table with RLE compression returned an ERROR stating that the physical datum size would exceed maximum data block size.
25835 Management Scripts: gptransfer 4.3.6.2 When the Greenplum Database gptransfer utility attempted to transfer data and the destination database did not contain a required filespace, the utility failed but did report that the destination database did not contain a required filespace.

Now, the utility reports the filespaces that are required in the destination database and exits.

25830 Query Execution 4.3.6.2 The legacy query optimizer returned incorrect results for some queries that contain a reference to a subquery that is defined in a WITH clause.

This issue has been resolved.

25819 Storage: Master Mirroring 4.3.6.2 When adding a standby master, the Greenplum Database gpinitstandby utility failed and caused a Greenplum Database postmaster reset if the utility could not copy a file from the primary master data directory to the standby master data directory.

This issue has been resolved. The utility reports an error and exits.

25799 DDL and Utility Statements 4.3.6.2 In some cases, altering a partitioned table with an ALTER TABLE command that contains both the clause SET WITH (REORGANIZE=false) and the clause DISTRIBUTED RANDOMLY caused a Greenplum Database PANIC.
25704 Storage: Vacuum 4.3.6.2 In some cases, a VACUUM FULL operation on an append-optimized table or column-oriented table would not vacuum the auxiliary tables associated the main table and would generate a notice stating that a relation could not be shrunk.

This issue has been resolved. The auxiliary tables are vacuumed if needed, and the notice does not appear.

25620 Backup and Restore 4.3.6.2 The Greenplum Database gpdbrestore utility returned an error when the -T and --ddboost options were specified to restore a single table from a Data Domain system with Data Domain Boost.
25617 Catalog and Metadata 4.3.6.2 In some cases, when DROP TABLE commands with the IF EXISTS clause were executed concurrently, an unnecessary ERROR was generated when Greenplum Database attempted to drop a table that was already dropped.

This issue has been resolved. The ERROR is no longer generated.

25549 Backup and Restore 4.3.6.2 Greenplum Database gpcrondump utility did not correctly back up a partitioned table when either the root parent partition table or child partition tables were altered to a different schema. This caused the Greenplum Database gpdbrestore utility to restore the partitioned table in schema of the root parent table.
20620 DDL and Utility Statements 4.3.6.2 Enhanced the Greenplum Database ALTER TABLE RENAME operation. The enhancement eliminates some possible catalog inconsistencies between the Greenplum Database master and segment instances.
25926 DDL and Utility Statements 4.3.6.1 When a user dropped a Greenplum Database table, the table was dropped only on the Greenplum Database master, not on the Greenplum Database segments, in this situation. The table is qualified by a schema name that is the same as the Greenplum Database user, the variable $user listed in the Greenplum Database search_path to include the user ID in the search path, and the DROP TABLE command issued by the user does not specify the table schema.

For example, the user is gpuser1, the table is gpuser1.table1, and the search_path contains "$user". The table table1 is dropped only on the Greenplum Database master when this command is run by gpuser1.

DROP TABLE table1 ;

This issue has been resolved.

This issue occurs with Greenplum Database 4.3.6.0. For instructions on how to check for and remove the orphan tables on the Greenplum Database segments, see Dropping Orphan Tables on Greenplum Database Segments.

25898 Storage: Transaction Management 4.3.6.1 Greenplum Database performance was degraded due to some Greenplum Database catalog queries that were hanging because Greenplum Database signal handlers did not properly stop the processes associated with the queries.

The ability of Greenplum Database signal handlers to stop processes has been enhanced.

25682 Management Scripts: gpssh 4.3.6.1 In some cases, the Greenplum Database utility gpssh failed to connect to some hosts specified either with the -h option or in the file specified by -f option. The utility successfully connected to other specified hosts.

This issue has been resolved.

25885 DDL and Utility Statements 4.3.6.1 After exchanging the default partition of a partitioned table with a table that contains data that violates the constraints defined on the default partition, some queries against the partitioned table returned incorrect results.

This issue has been resolved. For the ALTER TABLE command, the EXCHANGE DEFAULT PARTITION clause is no longer supported. For information about the ALTER TABLE command, see the Greenplum Database Reference Guide.

25865 Security 4.3.6.1 Greenplum Database software has been updated to resolve these security issues:
  • CVE-2015-0243. Memory errors in functions in the pgcrypto extension.
  • CVE-2015-0242. Buffer overrun in replacement printf family of functions.
  • CVE-2015-0241. Buffer overruns in "to_char" functions.
  • CVE-2013-1900. Random numbers generated by contrib/pgcrypto functions may be easy for another database user to guess.
25845 Backup and Restore 4.3.6.1 When restoring Greenplum Database database objects with the Greenplum Database utility gpdbrestore, tables for a specific schema could not specified.

This issue has been resolved. For information about the utility, see the Greenplum Database Utility Guide.

25839 Storage: Vacuum 4.3.6.1 For append-optimized tables that contain indexes, the VACUUM command scanned the table index to reclaim storage space from index data even if a vacuum operation was not performed the append-optimized table. This caused a performance issue.

This issue has been resolved. The VACUUM command performs a vacuum operation on indexes of append-optimized tables only when performing a vacuum operation on the append-optimized table

25830 Query Execution 4.3.6.1 When Pivotal Query Optimizer is enabled, some queries that contain the LEAST or GREATEST function and also contain a subquery that is defined inside the WITH clause might have fallen back to the Greenplum Database legacy optimizer and produced incorrect results.

This has been resolved. The query is executed by GPORCA and returns the correct results.

For Greenplum Database 4.3.6.2, the legacy optimizer returns correct results.

25806 Query Optimizer 4.3.6.1 For some queries, GPORCA was not able to use multi-column indexes because of incorrect metadata.

This issue has been resolved.

25794

25684

Catalog and Metadata 4.3.6.1 Greenplum Database requires a larger amount of memory during a Bitmap Index Scan depending on the number of rescans of the indexed relations. This memory requirement might cause some queries to run out of memory.

Memory management involving rescans of the indexed relations has been improved.

25787 Catalog and Metadata 4.3.6.1 If columns were dropped from a table and then ANALYZE was run on the table, the gp_toolkit view gp_stats_missing reported that the table did not have statistics.

This issue has been resolved.

25781 Backup and Restore 4.3.6.1 The Greenplum Database gpdbrestore utility failed when the -T or --table-file option was specified to restore database data if the data that was being restored contained a function definition that contains lines that begin with a SET command (without preceding spaces).

This issue has been resolved.

25744 Backup and Restore 4.3.6.1 When the Greenplum Database gpdbrestore utility was used to restore specific database tables from a backup on Data Domain server, other objects such as sequences were also incorrectly restored.

This issue has been resolved. For information about the utility, see the Greenplum Database Utility Guide.

25735 Upgrade / Downgrade 4.3.6.1 When upgrading from Greenplum Database 4.2.x, databases that contained 4.2.x append-only tables were not assigned the correct owner when they were converted to 4.3.x append optimized tables.

This issue has been resolved when upgrading from Greenplum Database 4.2.x to Greenplum Database 4.3.6.x. When upgrading from Greenplum Database 4.3.x for example, 4.3.5.3, to Greenplum Database 4.3.6.x, the 4.3.6.x installation contains an SQL script, fix_visimap_owner.sql, that fixes this ownership issue.

For information about using the script see Upgrading from 4.3.x to 4.3.6.x.

25724 Query Planner 4.3.6.1 For some queries that contained subqueries, the Greenplum Database planner generated an incorrect plan. When executing the plan, an error occurred when information from the outer query was not passed to the inner subquery. The error caused a Greenplum Database PANIC.
25710 DDL and Utility Statements 4.3.6.1 When default table storage options were specified with the Greenplum Database server configuration parameter gp_default_storage_options, the storage options were applied to Greenplum Database external tables.
25681 DDL and Utility Statements 4.3.6.1 When default table storage options were specified with the Greenplum Database server configuration parameter gp_default_storage_options, the storage options were applied to system tables.
25669 Backup and Restore 4.3.6.1 After an append-optimized table was restored with the Greenplum Database gpdbrestore utility, the table description was not restored.
25650 Storage: Segment Mirroring 4.3.6.1 Performance of the Greenplum Database gprecoverseg utility for incremental and full recovery of a Greenplum Database segment has been improved. For incremental resynchronization of segment mirrors, the Greenplum Database server configuration parameter filerep_mirrorvalidation_during_resync controls the of checking for the existence of files during incremental recovery.

For information about the utility, see the Greenplum Database Utility Guide. For information about the parameter, see the Greenplum Database Reference Guide.

25647 Storage: Access Methods 4.3.6.1 The processing of bit commits and abort transactions has been improved by eliminating look ups and contention on a global data structure during the operations.
25590 Interconnect 4.3.6.1 When no sockets were available on the Greenplum Database master, the message displayed was incorrect.
25589 Interconnect 4.3.6.1 When a Greenplum Database master instance failed over to the standby master, some processes that were controlled by the failed master were not shut down properly.
25588 Management Scripts: expansion 4.3.6.1 The Greenplum Database gpexpand utility performance has been enhanced.
25584 Query Execution 4.3.6.1 In some situations, a running Greenplum Database query could not be terminated with the functions pg_cancel_backend or pg_terminate_backend.

Now, pg_terminate_backend will cancel a running query.

25569 Management Scripts: gptransfer 4.3.6.1 When the Greenplum Database gptransfer utility --schema-only option was specified to create only database schemas in the target database, only one schema was created in the database. If the source database contained multiple schemas only one schema was created in the target database.
25564 Backup and Restore 4.3.6.1 When the Greenplum Database gpdbrestore utility options --redirect and --truncate were specified to restore database tables from a back up to a different database and then truncate the data from restored tables, the utility truncated the tables that were in the original database that was backed up.
25551 Loaders: Copy/ExternalTabs 4.3.6.1 In some cases, the Greenplum Database COPY command returned this error when copying data from a file that contained lines that used a mixture of newline types.
ERROR: COPY metadata not found.

This issue has been resolved.

25547 OS Abstraction 4.3.6.1 In some cases, Greenplum Database hung because the Greenplum Database logger process could not communicate with an SMTP server.

The SMTP communication software has been improved.

25516 DDL and Utility Statements 4.3.6.1 When shutting down a Greenplum Database, the Greenplum Database utility gpstop sometimes incorrectly returned a failure error code when it performed a forceful termination of Greenplum Database processes. This caused other Greenplum Database utilities that use gpstop, such as gpexpand, to fail because of the incorrect failure error code from gpstop.

The gpstop utility now returns the correct code.

25474 GPHDFS 4.3.6.1 When the Hadoop file system replication factor is 1, a Hadoop file created for an Greenplum Database external writable table with the gphdfs protocol was created with replication factor of 3.
25385 Client Access Methods and Tools 4.3.6.1 In some situations when Greenplum Database hung when sending data to a client application, the connection to the application could not be terminated with the pg_terminate_backend() function. The function has been improved.
25113 Storage: Access Methods 4.3.6.1 Greenplum Database did not manage the partition of partitioned tables efficiently. On a partitioned table with a large number of partitions, the frequent addition and removal of partitions caused the partitions ID numbers to incorrectly increase to more than the maximum value. This prohibited data from being added to the partition.

This issue has been resolved. Greenplum Database manages the partitions of partitioned tables more efficiently.

25081 Interconnect 4.3.6.1 In some cases, when a COPY command that contains a sub-select returns an error, Greenplum Database generated a segmentation fault.
24020 Backup and Restore 4.3.6.1 For the Greenplum Database pgcrondump utility, not all error messages were propagated from backup processes to utility during the backup process. This caused some failed backups to be reported as successful.
23963 Management Scripts: gptransfer 4.3.6.1 The Greenplum Database gptransfer utility would fail if it was run a second time after the gptransfer utility previously failed. The second failure occurred when the utility detected temporary database objects that were created in the target database when it previously failed.

This issue has been resolved. The gptransfer utility detects previously created temporary objects and deletes them before starting a transfer process.

23953

25662

25148

Management Scripts: gptransfer 4.3.6.1 The Greenplum Database gptransfer utility failed if the column data contained newline characters and in some cases when handling Null data and no data ('').

These issues have been resolved. The default gptransfer utility has been enhanced to use the CSV format to transfer data. The CSV format can handle data that contains newline characters and Null data.

For information about the utility, see the Greenplum Database Utility Guide.

4.3.5.x Releases

Table 28. Resolved Issues in 4.3.5.x
Issue Number Category Resolved In Description
25744 Backup and Restore 4.3.5.4 When the Greenplum Database gpdbrestore utility was used to restore specific database tables from a backup on Data Domain server, other objects such as sequences were also incorrectly restored.

This issue has been resolved.

25766 Query Optimizer 4.3.5.3 For some queries, GPORCA incorrectly converted a left outer join operation to an inner join operation during query optimization when a predicate contained a COALESCE function with constant arguments.

This issue has been resolved.

25743 Management Scripts: analyzedb 4.3.5.3 The Greenplum Database utility analyzedb returned an error if the utility encountered a table or schema that contained uppercase and lowercase characters.

This issue has been resolved.

25726 Query Optimizer 4.3.5.3 For some queries that have both outer joins and a predicate that contained COALESCE, GPORCA returned incorrect results.

This issue has been resolved.

25722 Query Optimizer 4.3.5.3 For some queries that contained computed columns in a subquery that were not used in the main query, Pivotal Query Optimizer generated an execution plan that contained the unused computed columns.

Now, Pivotal Query Optimizer generates a more efficient plan that does not contain unused computed columns.

25721 Query Optimizer 4.3.5.3 The Greenplum Database EXPLAIN command displays the setting of the server configuration parameter OPTIMIZER for a query and whether GPORCA or the legacy query optimizer generated the EXPLAIN plan.
25707 Query Execution 4.3.5.3 Pivotal Query Optimizer returned incorrect results for queries against partitioned tables that were partitioned by timestamp, and the query predicate for the partition selection was of type date.
25700 Query Optimizer 4.3.5.3 If the WHERE clause of a query compares a literal to a non-integer distribution column, and if the datatype of the literal differs from that of the distribution column, GPORCA might have generated a query execution plan that performed a dispatch to an incorrect segment. This dispatching problem caused incorrect results to be returned.

This issue has been resolved.

25697 Query Optimizer 4.3.5.3 For queries that contain a UNION or UNION ALL over multiple subexpressions, Pivotal Query Optimizer generated an execution plan with cascaded UNION or UNION ALL operators.

Now, a more efficient plan is generated with a single n-ary UNION or UNION ALL operator

25666 Query Execution 4.3.5.3 Greenplum Database supports INSERT operations into the leaf child partition table of a partitioned table.
24438 Query Optimizer 4.3.5.3 For queries that contain a GROUP BY clause that was used to group the results of a UNION or UNION ALL clause over more than two subexpressions, Pivotal Query Optimizer generated an execution plan that contained cascaded UNION or UNION ALL operators.

Now, the GROUP BY operation is pushed below the UNION or UNION ALL operator.

25667 Security 4.3.5.2 Greenplum Database software has been updated to use OpenSSL 0.9.8zg. For information about major changes in OpenSSL 0.9.8zg, see https://www.openssl.org/news/openssl-0.9.8-notes.html.
25661 Query Execution 4.3.5.2 Greenplum Database could not successfully execute some parametrized queries that included indexed columns where the indexed columns were queried but the parameters did not reference the indexed columns.

This issue has been resolved.

25643 Query Optimizer 4.3.5.2 For some queries that contained a computed column in a GROUP BY clause, Greenplum Database generated an execution plan that incorrectly pulled the computed column above the GROUP BY operation. This caused a Greenplum Database PANIC.

This issue has been resolved.

25637 Query Execution 4.3.5.2 Queries returned incorrect results if the query execution plan rescanned a bitmap index under a subplan where the bitmap was filtered using parameters from the outer plan.
25590 Interconnect 4.3.5.2 When no sockets were available on the Greenplum Database master, the message displayed was incorrect.
25589 Interconnect 4.3.5.2 When a Greenplum Database master instance failed over to the standby master, some processes that were controlled by the failed master were not shut down properly.
25588 Management Scripts: expansion 4.3.5.2 The Greenplum Database gpexpand utility performance has been enhanced.
25579 Query Optimizer 4.3.5.2 Some queries on partitioned tables caused a Greenplum Database PANIC if the query contains both of the following features:
  • An IN predicate that includes the table partitioning key
  • A subquery whose output column is the same partitioning key from the outer query

This issue has been resolved.

25576 Query Optimizer 4.3.5.2 For some star join queries with a large number of dimensions, the Pivotal Query Optimizer ran out of memory during the recursive processing of intermediate expressions.

This issue has been resolved by reducing the number of recursive steps performed on large expressions.

25516 Management Scripts: 4.3.5.2 When shutting down a Greenplum Database, the Greenplum Database utility gpstop sometimes incorrectly returned a failure error code when it performed a forceful termination of Greenplum Database processes. This caused other Greenplum Database utilities that use gpstop, such as gpexpand, to fail because of the incorrect failure error code from gpstop.

The gpstop utility now returns the correct code.

90922060 Query Execution 4.3.5.1 For queries where aggregates with distinct arguments are used as window functions, the query plan could have produced wrong results due to a limitation of the Window operator.

This issue has been resolved.

84334744 Query Optimizer 4.3.5.1 For some queries, an error occurs when the query contains a WITH clause (a common table expression) that references a column multiple times. In this example the column a is referenced multiple times:
WITH x AS (SELECT a AS a1, a AS a2 FROM t1) ...

This issue has been resolved.

25537 Query Execution 4.3.5.1 In some cases when GPORCA is enabled, a memory management issue occurred during a DynamicTableScan or DynamicIndexScan operation and caused a Greenplum Database PANIC.
25502 Storage: Access Methods 4.3.5.1 A Greenplum Database segment encountered an out of memory issue that caused a segmentation fault during the cleanup process. The segmentation fault caused a failover to a segment mirror.

The out of memory issue no longer results in a failover.

25490 Query Planner 4.3.5.1 Some SQL queries with nested subqueries returned this error:

ERROR: Failed to locate datatype for paramid 2

25485 Monitoring: gpperfmon server 4.3.5.1 If the password for the Greenplum Command Center user gpmon is not listed in the .pgpass file, the gpmmon process hung and users could not access Greenplum Command Center.
25484 DDL and Utility Statements, Storage: Transaction Management 4.3.5.1 Some queries were not handled properly by the Greenplum Database query dispatcher. This caused a PANIC on a Greenplum Database segment.
25476 Loaders: gpfdist 4.3.5.1 When running the Greenplum Database utility gpfdist, the size of log files increased quickly in some situations.

In Greenplum Database 4.3.5.1, the gpfdist utility supports the -s option to help minimize the increase of gpfdist log files. The -s option enables simplified logging. Only messages with WARN level and higher are written to the log file.

25472 Management Scripts: gpcheckcat 4.3.5.1 In some cases, the Greenplum Database utility gpcheckcat required a significant amount of time when it detected inconsistent OID (object ID) information and generated output about the inconsistencies.

In Greenplum Database 4.3.5.1, the gpcheckcat utility has been enhanced to reduce the time required when inconsistent OID information is detected. The utility creates verification files that contain a query that generates information about the inconsistent OIDs.

25464 Functions and Languages 4.3.5.1 Greenplum Database calculated the incorrect time for some queries due to an old version of a timezone file.

The Greenplum Database timezone files have been updated.

25455 Storage: Segment Mirroring 4.3.5.1 When running Greenplum Database utility gprecoverseg to perform an incremental segment recovery (the -F option was not specified), performance was poor if the database contained a large number of objects and a large number of DROP operations were performed.
The performance of gprecoverseg has been improved.
Note: The performance of gprecoverseg is not affected if the -F option is specified to perform a full segment recovery.
25421 Management Scripts: General 4.3.5.1 The Greenplum Database upgrade script fix_ao_upgrade.py failed on databases that do not contain the schema public.

This issue has been resolved.

25395 DDL and Utility Statements 4.3.5.1 When a Greenplum Database superuser issued a REVOKE command that did not cause a change the database, a warning was not issued.

Now, Greenplum Database displays a message that no privileges were revoked.

25386 Catalog and Metadata 4.3.5.1 The Greenplum Database utility gpcheckcat incorrectly issued messages for inconsistencies in the catalog for partitioned append-optimized tables.

The gpcheckcat utility no longer issues these messages.

25357 DDL and Utility Statements 4.3.5.1 In some cases, when running an ETL program against Greenplum Database, SQL DML statements such as SELECT, DELETE, INSERT, and UPDATE returned the warning unrecognized node type: 701.
25306 Backup and Restore 4.3.5.1 The Greenplum Database utility pg_dump did not back up aggregate functions correctly.
25158 Management Scripts: General 4.3.5.1 In some cases when Greenplum Database email alerts were enabled, the logger process crashed when trying to send an email. This resulted in fragmented pg_log files.

This issue has been resolved.

13685 Catalog and Metadata 4.3.5.1 In some cases, when an ALTER TABLE command that contained an EXCHANGE PARTITION clause was run to exchange table partitions, Greenplum Database did not use the same OID (object ID) for the related entry in the pg_constraint system catalog table on all the Greenplum Database segments.

The Greenplum Database gpcheckcat utility reported the inconstancy as an error when some database catalog checks were performed.

A consistent OID is now used.

11575 Catalog and Metadata 4.3.5.1 In some cases, when a CREATE INDEX command that contained a WHERE clause was run to create a partial index, Greenplum Database did not use the same OID (object ID) for the related entry in the pg_index system catalog table on all the Greenplum Database segments.

The Greenplum Database gpcheckcat utility reported the inconstancy as an error when some database catalog checks were performed.

A consistent OID is now used.

11289 Catalog and Metadata 4.3.5.1 In some cases, when performing CREATE TABLE or ALTER TABLE operations that include a default column value, Greenplum Database did not use the same OID (object ID) for the related entry the pg_attrdef system catalog table on all the Greenplum Database segments.

The Greenplum Database gpcheckcat utility reported the inconstancy as an error when some database catalog checks were performed.

A consistent OID is now used.

90561896 Management Scripts: recoverseg 4.3.5.0 In Greenplum Database 4.3.4.1, the Greenplum Database gprecoverseg utility checked persistent tables by default. In some cases, this check reported false positives for catalog corruptions.

For this release, the persistent table checks have been removed from gprecoverseg. To check for persistent table issues, use the Greenplum Database gpcheckcat utility.

The behavior in Greenplum Database has been reverted to the behavior in 4.3.4.0 and earlier.

89931274 Security 4.3.5.0 Greenplum Database software has been updated to use OpenSSL 0.9.8ze. For information about major changes in OpenSSL 0.9.8ze, see https://www.openssl.org/news/openssl-0.9.8-notes.html.
87808098 Loaders 4.3.5.0 In some cases, the Greenplum Database utility gpfdist issued an error message when a network connection to an ETL (extract, transform, and load) host was disconnected due to an issue with the host or the connection with the host.

The message has been enhanced to identify the cause of the error.

25423 Storage: Access Methods 4.3.5.0 In some cases, running the Greenplum Database utility gpcrondump caused a PANIC on some Greenplum Database segments.
25422 Management Scripts: General 4.3.5.0 The Greenplum Database utility gprecoverseg returns an error when the PGPORT environment variable is not set.
25417 Monitoring: gpperfmon server 4.3.5.0 In some cases when the Greenplum Command Center is installed, the gpsmon process timed out after one hour and returned an error that no request were received after 3600 seconds.
25339 Query Execution 4.3.5.0 In some cases, an out of memory error occurred during the evaluation of a per-row SQL function that required executing a nested query plan.
25335 Catalog and Metadata, Global Persistent Objects 4.3.5.0 In some cases, rebuilding a persistent table in a Greenplum database failed if the table was created with a non-default table space.
25311

25350

Query Optimizer 4.3.5.0 During query optimization, some queries with a large number of conjunctive predicates could consume a large amount of memory.

This issue has been resolved.

25305 Backup and Restore 4.3.5.0 The Greenplum Database utility gpmfr failed when all the files that were being backed up were less than 1000 bytes.
25297 Query Optimizer 4.3.5.0 Some queries returned wrong results when an IN clause contained a nested expression.

This issue has been resolved.

25296 Query Optimizer 4.3.5.0 Some queries that contained aggregate functions were terminated by an error because of incorrect plans within a window function.

This issue has been resolved.

25292

25361

Query Optimizer 4.3.5.0 Inefficient plans were generated for queries that contained the function unnest(ARRAY[...]). The plan generated by the Pivotal Query Optimizer for this type query has been improved.
25288 Query Execution 4.3.5.0 A Greenplum Database PANIC occurred when deleting data from a table with the DELETE command if the contains a USING clause.
25279 Management Scripts: gpstart/gpstop 4.3.5.0 In some cases, the Greenplum Database utility gpstop issued the warning No leftover gpmmon process found. These warning messages have be changed to informational messages.
25252 Query Optimizer 4.3.5.0 Some queries that required partition elimination with a NOT IN predicate caused a crash.

This issue has been resolved.

25175 Query Execution 4.3.5.0 A Greenplum Database PANIC occurred when using the COPY command to copy data into a table that contains no columns.
25170 Storage: Vacuum 4.3.5.0 In some cases, running the Greenplum Database utility vacuumdb caused a Greenplum Database PANIC due to issues with a system table that is used to track append optimized file segments.
25160 Query Execution 4.3.5.0 In some cases, running a query and concurrently performing a DDL operation on the same data returned this error.
ERROR", "XX000","could not open relation
25124 Dispatch 4.3.5.0 If a cursor was declared in a transaction, and then a SET command was issued in the same transaction before the cursor was closed, a Greenplum Database crash occurred.

In Greenplum Database 4.3.5.0, an error is returned if the SET command is issued while a cursor is declared and not closed in a transaction. In the transaction, the cursor must be closed before the SET command can be issued.

25081 Interconnect 4.3.5.0 In some cases, when a COPY command that contains a sub-select returns an error, Greenplum Database generated a segmentation fault.
24953 Management Scripts: gptoolkit 4.3.5.0 For append-optimized tables, display information about the amount of bloat (table disk space is occupied by deleted or obsolete rows) in the on-disk data files that are used by the tables.

The Greenplum Database function __gp_aovisimap_compaction_info displays append-optimized table on-disk storage and bloat information. See "The gp_toolkit Administrative Schema" in the Greenplum Database Reference Guide.

24944 DDL and Utility Statements 4.3.5.0 The set_config() function changed the sever configuration parameter only on the Greenplum Database master, not on the Greenplum Database segment instances.
24621 Backup and Restore, Functions and Languages 4.3.5.0 The Greenplum Database function to_date() did not validate the range of the input date.
24591 Backup and Restore 4.3.5.0 In some cases, the Greenplum Database utility gpcrondump failed with the error Cannot allocate memory.

The memory management of the Greenplum Database utility has been enhanced to minimize occurrence of the error.

24557 Query Optimizer 4.3.5.0 Some queries with aggregate functions that contained outer references returned the error message: aggref found in non-Agg plan node.

This issue has been resolved.

24263 Query Optimizer 4.3.5.0 Some queries with predicates on the join key of a left outer join did not push down a predicate. The plan generated by GPORCA for this type query has been improved.
23801 DDL and Utility Statements 4.3.5.0 For a table with a primary key, the ALTER TABLE command could change the distribution policy (the columns specified with the DISTRIBUTION KEY clause) to a non-primary key. Specifying a DISTRIBUTION KEY to a non-primary key column is not supported.

This issue has been resolved.

18673 DDL and Utility Statements 4.3.5.0 In some cases, SQL commands that were executed concurrently with an ALTER TABLE command that contains a SPLIT PARTITION clause on a partitioned table returned this error:
ERROR: Relation decrement reference count
   found relation relation-id with bad count

4.3.4.x Releases

Table 29. Resolved Issues in 4.3.4.x
Issue Number Category Resolved In Description
25422 Management Scripts: General 4.3.4.2 The Greenplum Database utility gprecoverseg returned an error when the PGPORT environment variable was not set.
25417 Monitoring: gpperfmon server 4.3.4.2 In some cases when the Greenplum Command Center is installed, the gpsmon process timed out after one hour and returned an error that no request were received after 3600 seconds.
0561896 Management Scripts: recoverseg 4.3.4.2 In Greenplum Database 4.3.4.1, the Greenplum Database gprecoverseg utility checked persistent tables by default. In some cases, this check reported false positives for catalog corruptions.

For this release, the persistent table checks have been removed from gprecoverseg. To check for persistent table issues, use the Greenplum Database gpcheckcat utility.

The behavior in Greenplum Database 4.3.4.2 has been reverted to the behavior in 4.3.4.0 and earlier.

25202 Management Scripts: gpstart/gpstop 4.3.4.1 The Greenplum Database utility gpstop did not clean up shared memory after stopping segments.
25201 Storage: Access Methods 4.3.4.1 A Greenplum Database PANIC occurred when altering a column and adding a new column in the same command on an append-optimized column-oriented table.
25193 Query Execution 4.3.4.1 In some cases, a PL/pgSQL exception handler disabled sequence access for a query. This caused Greenplum Database to return an error if the same query required a new sequence value after the PL/pgSQL exception.
25192 Storage: Transaction Management 4.3.4.1 For some queries that create sub-transactions, a fix in an earlier release degraded performance.

The performance degradation issue has been resolved.

25162 Storage: Access Methods 4.3.4.1 In previous releases, the Greenplum Database server configuration parameter gp_appendonly_compaction_threshold was not visible. This issue has been resolved.

For information about the parameter, see "Server Configuration Parameters" in the Greenplum Database Reference Guide.

25082 Languages: R PLR 4.3.4.1 The Greenplum Database utility gpssh supports the -s option to source the file greenplum_path.sh before executing any commands on the target host.

For information about the utility, see the Greenplum Database Utility Guide.

25066 Storage: Access Methods 4.3.4.1 In some cases, a Greenplum Database PANIC occurred when a query performed an index scan on an append-optimized table that was populated with concurrent inserts followed by updates or deletes.
25029 Upgrade / Downgrade 4.3.4.1 In some cases, the Greenplum Database utility gpmigrator_mirror failed when upgrading from 4.2.x to 4.3.x.
25017 Monitoring: gpperfmon server 4.3.4.1 In some cases, high memory consumption spikes were caused by gpmon. The memory consumption spikes might have caused gpmmon segmentation faults.

The Greenplum Database Command Center parameter ignore_qexec_packet reduces the spikes of memory consumption. See the Greenplum Command Center Administrator Guide.

25004 Upgrade / Downgrade 4.3.4.1 In Greenplum Database 4.3.3 and later, the GRANT and REVOKE commands support the TRUNCATE privilege on a table. The upgrade process to a release between 4.3.3 and 4.3.4 did not grant the TRUNCATE permission to database tables that were created before the upgrade.

This issue has been resolved when upgrading to 4.3.4.1.

24964 Management Scripts: recoverseg 4.3.4.1 The Greenplum Database utility gprecoverseg did not ensure segments were stopped before attempting segment recovery.
24899 Functions and Languages 4.3.4.1 The aggregation function sum failed when the function arguments are empty arrays.
24898 Management Scripts: recoverseg 4.3.4.1 The Greenplum Database utility gprecoverseg failed when Greenplum Database persistent tables contained inconsistencies.

Now, before attempting a segment recovery, the utility checks for persistent table inconsistencies and returns an error when inconsistencies are found.

24841 Storage: Access Methods 4.3.4.1 Running an UPDATE command after a DROP COLUMN and ADD PARTITION command on a partitioned table caused a Greenplum Database segment instance failure.
24740 Storage: Segment Mirroring 4.3.4.1 In some cases, interrupts that occurred during Greenplum Database query processing were not handled properly. This caused an increase in CPU utilization and a decrease in performance.
24678 Storage: Segment Mirroring 4.3.4.1 A performance issue was found with Greenplum Database during table scans. In some cases, unnecessary network traffic and increased disk activity occurred during read operations during the table scan.
24632 Catalog and Metadata 4.3.4.1 A SELECT command failed on a table if the table was defined with DISTRIBUTED BY clause that contained more than 200 columns.

This issue has been resolved.

24563

20470

Query Planner 4.3.4.1 For some queries that required a correlated subquery, Greenplum Database generated a query plan that caused a segmentation fault when the plan was executed.

This issue has been resolved.

24505 Management Scripts: recoverseg 4.3.4.1 The Greenplum Database utility gprecoverseg did not properly handle a segment that was in a transition phase, for example the primary segment is down and the mirror segment has not yet become the primary.

Now the utility retries segment recovery in this situation.

23267 Functions and Languages 4.3.4.1 Memory leaks in the Greenplum Database aggregate functions string_agg and array_agg have been fixed.
21724 Query Planner 4.3.4.1 Greenplum Database executes an SQL query in two stages if a scalar subquery is involved. The output of the first stage plan is fed into the second stage plan as a external parameter. If the first stage plan generated zero tuples and directly contributed to the output of the second stage plan, incorrect results might have been returned.

This issue has been resolved.

17709 Catalog and Metadata 4.3.4.1 When the Greenplum Database utility gpcheckcat was run with the -A option to check all databases, the utility overwrote repair scripts for each database.

This issue has been resolved.

24913 Storage Access Methods 4.3.4 In some cases, after a VACUUM operation was performed on an append-optimized table that contains an index, SQL queries that were run against the table returned incorrect data.
24778 Security 4.3.4 Greenplum Database software has been updated to use OpenSSL 0.9.8zc in response to the OpenSSL Security Advisory [15 Oct 2014]. For information about the advisory, see https://www.openssl.org/news/secadv_20141015.txt
24770 Dispatch 4.3.4 In some cases, NULL pointer handling was not performed correctly during query execution. This caused a panic on Greenplum Database master.
24694 Upgrade / Downgrade 4.3.4 The attribute name of a visimap metadata table for append-optimized tables was incorrectly named during the upgrade to 4.3.x from Greenplum Database 4.2.x.

This issue has been resolved.

24693   4.3.4 For the visimap metadata table for append-optimized tables, the value of attstorage was set incorrectly in the pg_attribute table during the upgrade to 4.3.x from Greenplum Database 4.2.x.

This issue has been resolved.

24692 Upgrade / Downgrade 4.3.4 A unnecessary record for the internal built-in function pg_get_stat_walsenders was added to the pg_depend table during the upgrade to 4.3.x from Greenplum Database 4.2.x.

This issue has been resolved.

24684 Locking, Signals, Processes 4.3.4 In some cases, Greenplum Database did not clean up processes correctly during a commit transaction. This caused crash recovery issues after a Greenplum Database restart of both a primary segment and its mirror segment.
24683 Backup and Restore 4.3.4 There were performance issues when backing up specific tables from Greenplum Database using the Greenplum Database utility gpcrondump with the --table-file, -t, or --exclude-table-file, -T options.

The back up performance has been improved.

24679 Locking, Signals, Processes 4.3.4 In some cases, Greenplum Database PL/Java did not exit properly and caused crash recovery issues after a Greenplum Database restart of both a primary segment and its mirror segment.
24677 Backup and Restore 4.3.4 In some cases after a successful back up operation, error messages about lock files were incorrectly displayed.

This issue has been resolved.

24667 Build, Test 4.3.4 When creating a temporary table with the ON COMMIT DELETE ROWS clause in a heavy workload environment, the prepared transaction that created the temporary table failed in some cases.
24662 Upgrade/Downgrade 4.3.4 The privilege of the internal system view pg_stat_walsenders was set to NULL during the upgrade to 4.3.x from Greenplum Database 4.2.x.

The upgrade process now sets the privilege correctly.

24630 Management Scripts: gpsd 4.3.4 The Greenplum Database support utility gpsd has been enhanced to be more memory efficient in order to prevent failures when used on a database that contains large schemas.
24606 Storage access method 4.3.4 In some cases, queries that were run against the gp_toolkit view gp_size_of_table_and_indexes_disk returned an error when the queries are run concurrently with DDL statements.
24515 Replication: Segment Mirroring 4.3.4 In some cases under a heavy workload, logging onto the Greenplum Database segment host as a UNIX user was not possible. This was caused by a Greenplum Database filerep process that was incorrectly sending signals to the user after the process failed to create a sub-process.
24383 GPHDFS 4.3.4 Greenplum Database external tables did not support using the gphdfs protocol and MapR to access HDFS data.
24264 Catalog and Metadata 4.3.4 The commands REINDEX TABLE table_name and REINDEX INDEX index_name did not re-index child partition indexes of a partitioned table.
24216 Management Scripts: gpstart/gpstop 4.3.4 In some cases, the Greenplum Database utility gpstop failed to shutdown a database.
24052 Query Planner 4.3.4 In some cases, when executing a query that joined tables and the tables have extremely inaccurate statistics, Greenplum Database would return an error that reported being out of memory.

This issue has been resolved.

24003 Dispatch, Interconnect 4.3.4 In some cases, executing an SQL query caused a memory leak.
23802 Query Execution 4.3.4 Greenplum Database did not manage temporary workfiles (spill files) properly. In some cases, this caused a query that required workfiles to fail with a message that stated that a Greenplum Database segment had reached the maximum configured workfile usage limit.
23751 Monitoring: gpperfmon server 4.3.4 In some cases, a memory leak caused the gpmmon process to consume a large amount of memory and CPU resources.
23395 Transaction Management 4.3.4 If a Greenplum Database segment failed during two phase transaction processing, the transaction remained in a uncompleted state and was cleaned up only during a Greenplum Database restart. In many cases, this caused high disk consumption by the Greenplum Database xlog process.
23130 Resource Management 4.3.4 Some queries were terminated when they were run with a specific statement_mem value due to a rounding error.
23031 Backup and Restore 4.3.4 When restoring a specific table that was not in the default schema (public) with the gpdbrestore utility from a back up, the table was restored in the public schema.
22526 OS Abstraction 4.3.4 In some cases, Greenplum Database would wait indefinitely for a response from a client application.

This issue has been resolved. See the server configuration parameter gp_connection_send_timeout.

19612 Replication: Segment Mirroring 4.3.4 In some cases, Greenplum Database did not properly handle the transition when segment mirror becomes unavailable. As a result, Greenplum Database becomes unavailable.
18509 Functions and Languages 4.3.4 In some cases, Greenplum Database did not handle data of type date properly and caused a segmentation fault.

4.3.3.x Releases

Table 30. Resolved Issues in 4.3.3.x
Issue Number Category Resolved In Description
24677 Backup and Restore 4.3.3.1 In some cases after a successful back up operation, error messages about lock files were incorrectly displayed.

This issue has been resolved.

24667 DDL and Utility Statements 4.3.3.1 When creating a temporary table with the ON COMMIT DELETE ROWS clause in a heavy workload environment, the prepared transaction that created the temporary table failed in some cases.
24641

24508

Monitoring: Commander Center Alerting 4.3.3.1 In the Greenplum Command Center database gpperfmon, performance issues were caused by data skew in the log_alert_history. The distribution key for the table has been changed to resolve the issue. See Changed Features.
24515 Replication: Segment Mirroring 4.3.3.1 In some cases under a heavy workload, logging onto the Greenplum Database segment host as a UNIX user was not possible. This was caused by a Greenplum Database filerep process that was incorrectly sending signals to the user after the process failed to create a sub-process.
23751 Monitoring: gpperfmon server 4.3.3.1 In some cases, a memory leak caused the gpmmon process to consume a large amount of memory and CPU resources.
23395 Transaction Management 4.3.3.1

If a Greenplum Database segment failed during two-phase transaction processing, the transaction remained in a uncompleted state and was cleaned up only during a Greenplum Database restart. In many cases, this caused high disk consumption by the Greenplum Database xlog process.

18509 Functions and Languages 4.3.3.1 In some cases, Greenplum Database did not handle data of type date properly and caused a segmentation fault.
24479 Backup and Restore 4.3.3.0 A table could not be restored (with the gpdbrestore -T option) from a back up that is on a Data Domain Boost system and that was created with the gpcrondump --ddboost options.
24478 Management Scripts: expansion 4.3.3.0 The Greenplum Database gpexpand utility failed when an error table for an external table was present in Greenplum Database. The utility displayed this message:

DETAIL: ALTER TABLE is not allowed on error tables

24326 Query Execution, Storage Access Methods 4.3.3.0 If either a non-partitioned append-only table or an individual append-only part of a partitioned table had more than 127 million rows on a segment, a query that uses an index to access the table data could return duplicate rows.

This issue has been fixed.

24317 Security 4.3.3.0 Greenplum Database software has been updated to use OpenSSL 0.9.8zb in response to the OpenSSL Security Advisory [6 Aug 2014]. For information about the advisory, see https://www.openssl.org/news/secadv_20140806.txt.
24248 GPHDFS 4.3.3.0 The Greenplum Database external table protocol gphdfs supports the Cloudera 4.x and 5.x HDFS distributions. See Hadoop Distribution Compatibility
24237 DDL and Utility Statements 4.3.3.0 Temporary tables were not cleaned up properly in the following situation. A user defined function (UDF) was created with a security definer and that included statements to create the temporary table. The UDF was executed by a regular user who was given EXECUTE permission on the function.

This caused the temporary table to stay in the database after the session was disconnected.

24182 Management Scripts: General 4.3.3.0 Greenplum Database timezone information has been updated to match world-wide timezones. For in formation about timezones, see http://www.iana.org/time-zones.
24168 Vacuum 4.3.3.0 For an append-optimized table that did not contain any data, the VACUUM command did not update the value of relfrozenxid in the catalog table pg_class.
24158 Upgrade / Downgrade 4.3.3.0 When upgrading Greenplum Database from a 4.2.x release to a 4.3.x release prior to 4.3.2.1, append-only tables were not correctly converted to append-optimized tables. In some cases, the incorrect conversion prevented the VACUUM command from reclaiming storage occupied by deleted tuples. For information about the upgrade issue, see Upgrading from 4.3.x to 4.3.3.x.
24119 Query Execution 4.3.3.0 In some cases, a segmentation fault occurred when a DECLARE CURSOR WITH HOLD command was run by an ODBC driver.
24116

21042

Loaders: gpfdist 4.3.3.0 The Greenplum Database gpfdist utility failed with a SIGSEGV error when the utility received a empty request with two consecutive return characters "\n\n".
24089 Loaders: Copy/ExternalTabs 4.3.3.0 Multibyte characters were not handled properly when writing to an external table that uses the gb18030 encoding from a Greenplum database that was created with UTF8 encoding. In some cases, this error was encountered.

ERROR: The size of the value cannot be bigger than the field size value

24079 GPHDFS 4.3.3.0 The Greenplum Database external table protocol gphdfs supports the Pivotal 2.0 distribution.
24068 Postgis 4.3.3.0 When using PostGIS, In some cases a closed curved polygon that was converted to a linear polygon was not closed due to a linear approximation precision issue with PostGIS 2.0.3.
24067 Loaders: gpfdist, Loaders: gpload 4.3.3.0 In some cases when network load was heavy, the Greenplum Database utility gpfdist intermittently failed with this error: gpfdist closed connection to server
24055 Vacuum 4.3.3.0 The VACUUM FULL command transaction processing has been enhanced ensure proper operation with other concurrent operations.
24011 Catalog and Metadata, Vacuum 4.3.3.0 In some cases, when a VACUUM FULL command was cancelled, incorrect handling of Greenplum Database transaction log caused a PANIC signal to be issued and prevented Greenplum Database from performing a crash recovery of a segment mirror.
24001 Backup and Restore 4.3.3.0 During a backup operation, the Greenplum Database utility gpcrondump held an EXCLUSIVE lock on the catalog table pg_class longer than required.
23955 Query Execution 4.3.3.0 In some query plans, where a window operator is under the right child of a nested loops join, wrong results could have been generated because of improper cleanup of the operator's internal state.
23925 Management Scripts: expansion, Management Scripts: General 4.3.3.0 The Greenplum Database utilities gpactivatestandby and gpexpand used SSH to connect to localhost (the Greenplum Database host where the utility was run). Using SSH was redundant as the command was already on the local host and has been eliminated.
23894 Backup and Restore 4.3.3.0 Performing a back up to a Data Domain system failed when the Greenplum Database utility gpcrondump command specified the --ddboost options because gpcrondump performed a disk space check.
23864 Catalog and Metadata 4.3.3.0 Running the REINDEX command on a database while other workloads are concurrently running could create inconsistencies in the database catalog.
23850 Management Scripts 4.3.3.0 In some cases after expanding a Greenplum Database system, running gpinitstandby -n failed to resynchronize the data between the primary and standby master host.
23850 Management Scripts: General 4.3.3.0 In some cases after expanding a Greenplum Database system, running gpinitstandby -n failed to resynchronize the data between the primary and standby master host.
23842 Replication: Segment Mirroring 4.3.3.0 In some rare cases, if a restart occurred while the gprecoverseg utility was running, some tables and a persistent table were detected having less data on a mirror segment that corresponds to a primary segment.
23802 Query Execution 4.3.3.0 Greenplum Database did not manage temporary workfiles (spill files) properly. In some cases, this caused a query that required workfiles to fail with a message that stated that a Greenplum Database segment had reached the maximum configured workfile usage limit.
23753 Backup and Restore 4.3.3.0 The emails sent by the Greenplum Database gpcrondump utility could not be customized. Now the utility supports customized email notification for backup operations.
23730 Backup and Restore, Management Scripts: master mirroring 4.3.3.0 When configuring a Greenplum Database system with a standby master, the gpinitstandby utility did not correctly update the pg_hba.conf file on Greenplum Database segment hosts.
23729 Backup and Restore, DDL and Utility Statements 4.3.3.0 When the -b option was specified with the gpcrondump utility to disable a disk space check, a check was still performed.
23717 Locking, Signals, Processes 4.3.3.0 During Greenplum Database shutdown, a signal-unsafe function call was called from a signal handler function. The signal-unsafe function was replaced.
23699 Monitoring: gpperfmon server 4.3.3.0 Greenplum Database failed when the gpperfmon log files were not encoded in UTF8.

This issue has been resolved.

23637 Backup and Restore 4.3.3.0 When restoring a Greenplum database with the Greenplum Database gpcrondump utility, the utility performed an ANALYZE operation on the entire database. Now the gpcrondump utility analyzes only the restored tables.
23568 Backup and Restore 4.3.3.0 When backing up a Greenplum database with the Greenplum Database gpcrondump utility and specifying an NFS directory with the -u option, the gpcrondump utility created an empty db_dumps directory in the master and segment data directories.
23558 Backup and Restore 4.3.3.0 When restoring a backup from a Data Domain system using --ddboost options, the Greenplum Database gpdbrestore utility failed because it could not find C data and post data files.
23286 Dispatch 4.3.3.0 In some cases, Greenplum Database did not handle the processing of cancelled distributed queries properly.

This issue has been resolved.

22974 Loaders: Copy/ExternalTabs 4.3.3.0 When reading data from external sources, Greenplum Database stopped reading data if the first 1000 rows processed contain formatting errors. Now the limit is configurable.
20504 Query Execution 4.3.3.0 FOR loops in PL/pgSQL did not close the sequence generator if further access was still required.
18562 DDL and Utility Statements 4.3.3.0 A transaction lock did not block reader processes from proceeding when a writer process was holding the same lock. In some cases this caused a race condition to occur.

Now, Greenplum Database blocks reader processes when a writer process holds the same lock to prevent race conditions from occurring.

17264 Replication: Segment Mirroring 4.3.3.0 In some cases, Greenplum Database continuously logged this message when sending file replication process statistics to Greenplum Database perfmon process:

Error when sending file rep stats to perfmon

16450 Backup and Restore 4.3.3.0 When running the Greenplum Database utility pg_dumpall with the option --resource-queues to create scripts that contain resource queue definitions, the utility generated incorrect scripts when the resource queue definition contained the memory_limit option.
16059 Resource Management 4.3.3.0 Some SQL statements that executed a PL/pgSQL function that contained an insert, update, or delete operation did not allocate memory correctly. This caused the following issues:
  • Decreased performance
  • Large number of spill files were created on disk

This issue has been resolved.

4.3.2.x Releases

Table 31. Resolved Issues in 4.3.2.x
Issue Number Category Resolved In Description
24158 Upgrade / Downgrade 4.3.2.1 When upgrading Greenplum Database from a 4.2.x release to a 4.3.x release prior to 4.3.2.1, append-only tables were not correctly converted to append-optimized tables. In some cases, the incorrect conversion prevented the VACUUM command from reclaiming storage occupied by deleted tuples. For information about the upgrade issue, see Product Enhancements.
24326 Query Execution, Storage Access Methods 4.3.2.1 If either a non-partitioned append-only table or an individual append-only part of a partitioned table had more than 127 million rows on a segment, a query that uses an index to access the table data could return duplicate rows.

This issue has been fixed.

24037 Client Access Methods and Tools 4.3.2 In some cases, when the SQLCancel function was used with the Greenplum Database ODBC driver to cancel the execution of a query, a rollback of the transaction did not occur.
23838 Loaders: Copy/External Tables 4.3.2 When the COPY command copied data from a file and the file contained the character sequence '\r\r\n', a postmaster reset occured.
23768 Query Execution 4.3.2 In some cases, the clean up of an aborted transaction was not handled correctly and caused a PANIC signal to be issued.
23751 Monitoring: gpperfmon server 4.3.2 A memory leak caused the gpmmon process to consume a large amount of memory and CPU resources.
23735 Languages: PL/Java 4.3.2 In some cases, Greenplum Database did not handle concurrent shared memory operations properly from PL/Java routines. This caused a PANIC signal to be issued.
23708 Backup and Restore 4.3.2 In some cases, running the Greenplum Database gpdbrestore utility with the -T or --table-file option failed with this error.

ValueError: need more than 1 value to unpack

23706 Upgrade / Downgrade 4.3.2 The Greenplum Database installer did not support upgrading from a Greenplum Database hotfix.
23647 Vacuum 4.3.2 Performing a VACUUM operation on a partitioned append-optimized table did not correctly reduce the age of the parent table and child tables.
23631 Replication: Segment Mirroring 4.3.2 In some rare cases, the crash recovery of a segment mirror failed due to an inconsistent LSN.
23604 Interconnect 4.3.2 In some cases when a Greenplum Database process was cancelled on the Greenplum Database master, corresponding processes remained running on Greenplum Database segment instances.
23578 gphdfs 4.3.2 For Greenplum Database external tables, the gphdfs protocol that accesses data from files on HDFS now supports the CSV file format.
23546 Storage Access Methods 4.3.2 In some cases, a DELETE command that contains a join between an append-optimized table and heap table returned this error:

ERROR: tuple already updated by self

23485 Transaction Management 4.3.2 When a single Greenplum Database session ran transactions, temporary files were not removed after the transaction completed. If a the session ran a large number of transactions, the temporary files required a large amount of disk space.

This issue has been resolved.

23417 Transaction Management 4.3.2 Some queries against an append-optimized table with compression enabled that containd a column with an unknown data type caused a Greenplum Database SIGSEGV error.
23227 Client Access Methods and Tools 4.3.2 For Greenplum Database with GSS Authentication enabled, the database role attribute Valid Until was ignored. The Valid Until parameter is now respected when GSS authentication is enabled.
23222 Client Access Methods and Tools 4.3.2 When Greenplum Database receives a SIGSEGV when running the COPY command, Greenplum Database hangs and continuously log this warning message:

copy: unexpected response (3)

23204 Query Execution 4.3.2 In some cases, when a Greenplum Database segment fault occurred during the execution of a PL/R function, PL/R hung and continuously returned the same error message.
23202 Management Scripts: expansion 4.3.2 During the process of adding new hosts, the Greenplum Database expand utility gpexpand did not update the pg_hba.conf files on Greenplum Database hosts with the correct host information.
23174 Languages: R, PLR 4.3.2 In Greenplum Database, a signal handling issue in the R programming language caused a potential for postgres processes to hang when running PL/R functions.
23138 Replication: Segment Mirroring 4.3.2 The gprecoverseg utility failed to recover a Greenplum Database segment that was marked as down when the data directory location for the segment was a symbolic link, and a postgres process was running with the same PID as the PID associated with the down segment.
23067 Loaders: Copy/External Tables 4.3.2 In some cases, when an INSERT FROM SELECT command was run that selected from readable external able and inserted into writable external table, this warning was generated:

WARNING select failed on curl_multi_fdset (maxfd 10) (4 - Interrupted system call)

23038 Query Execution 4.3.2 When a query was run that contained a polymorphic, user-defined aggregate function, and Greenplum Database was required to create spill files on disk, the query failed with this error.

ERROR: could not determine actual argument type for polymorphic function

This issue has been fixed.

23008 Dispatch 4.3.2 In some cases when temporary tables were used, Greenplum Database did not perform the clean up of temporary namespaces properly after a transaction completed and caused a SIGSEGV.
22914 Loaders: Copy/ExternalTables 4.3.2 When a query joined a heap table with an external table that used the gpfdist protocol, an incorrect plan that returned no results might have been chosen.
22787 Monitoring: gpperfmon server 4.3.2 In some cases, the Greenplum Database gpmmon process failed. The gpmmon process is used for Greenplum Database performance monitoring.
22784 Storage Access Methods 4.3.2 After a database expansion, some tables created with APPENDONLY=TRUE and compression enabled consumed much more disk space than before the expansion.

To reduce disk space in this situation, the Greenplum Database gpreload utility reloads table data with column data sorted.

22706 Management Scripts: master mirroring 4.3.2 The Greenplum Database gpinitstandby utility completed successfully but returned an error when the $GPHOME/share directory was not writable. Now, the utility returns this warning:

Please run gppkg --clean after successful standby initialization.

22592 Backup and Restore 4.3.2 When the Greenplum Database gpdbrestore utility could not find files on the Greenplum Database master segment that are used to perform a restore operation, the utility did not return the correct error message.
22413 Query Planner 4.3.2 In some cases, an SQL query that contains the following returned incorrect results: a combination of a median function with other aggregates where the GROUP BY columns are a subset of the table's distribution columns.
22328 Management Scripts 4.3.2 When a Greenplum Database extension package was updated with the Greenplum Database gppkg utility option -u, gppkg did not warn the user that updating a package includes removing all previous versions of the system objects related to the package.

Now, the gppkg utility warns the user and lets the user cancel the operation.

22265 Locking, Signals, Processes 4.3.2 Greenplum Database hung due to incorrect lock handling that caused a race condition. The lock handling issue was caused by a compiler optimization.
22205 Replication: Segment Mirroring 4.3.2 In some cases, running the Greenplum Database command gprecoverseg -r to rebalance segment instances failed and caused database catalog corruption.
21916 Interconnect 4.3.2 In some cases when the Greenplum Database query dispatcher encountered connection errors, a postmaster reset occured.
21867 DDL and Utility Statements 4.3.2 The performance of Greenplum Database truncate operations degraded between restarts of Greenplum Database.
21103 Query Execution 4.3.2 In Greenplum Database, support of subnormal double-precision (float8) numbers differed between Red Hat Enterprise Linux 5 and Red Hat Enterprise Linux 6. For example, the value 5e-309 was not handled consistently by Greenplum Database on RHEL 5 and RHEL 6. This issue has been resolved.
20600 Query Planner 4.3.2 For some SQL queries that contained a subquery, this error message was returned. ERROR: no parameter found for initplan subquery.
20268 Loaders: Copy/ExternalTables 4.3.2 In some cases when a COPY command was run, improper memory handling caused a PANIC signal to be issued.
19949 Backup and Restore 4.3.2 If a Greenplum database was backed up and the database name contained upper-case characters, the Greenplum Database gpdbrestore utility did not restore the database with the correct name.
19660 Authentication 4.3.2 Greenplum Database supports LDAP authentication. Previously, an issue in Greenplum Database prevented LDAPS (LDAP over SSL) from functioning. This issue has been resolved.
19246 Backup and Restore 4.3.2 When performing a selective restore of a partitioned table from a full backup with the Greenplum Database utility gpdbrestore, the data from leaf partitions are now restored.

Previously, when performing a selective restore of a partitioned table, you needed to specify all the individual leaf partitions.

18774 Loaders 4.3.2 External web tables that use IPv6 addresses no longer require a port number when using the default port.

In previous releases, a port number was required when using an IPv6 address.

13282 Backup and Restore 4.3.2 The database objects in the gp_toolkit schema were not restored after a database was re-created and then restored with the Greenplum Database gpdbrestore utility. The gp_toolkit objects are now restored when a database is re-created and restored.

4.3.1 Release

Table 32. Resolved Issues in 4.3.1
Issue Number Category Resolved In Description
23757 Security 4.3.1 Greenplum Database software has been updated to use OpenSSL 0.9.8za in response to the OpenSSL Security Advisory [05 Jun 2014]. For information about the advisory, see https://www.openssl.org/news/secadv_20140605.txt.
22301 Replication: Master Mirroring 4.3.1 DCA customers who wished to use Greenplum Database 4.3 could not use the utility dca_setup. This issue has been resolved in Greenplum Database 4.3.1.
22281 Backup and Restore 4.3.1 For partitioned append-optimized tables, a partition was backed up even though it was not modified.
21591 Management Scripts Suite 4.3.1 The Greenplum Database utilities gpstart and gprecoverseg hung when checking the process ID in the postmaster.pid file and the ID matched a non-postgres running process.
23421 Locking, Signals, Processes 4.3.1 In some cases, concurrent CREATE TABLE and DROP TABLE operations caused Greenplum Database to hang due to incorrect lock handling.
13825 Functions and Languages, Transaction Management 4.3.1 In PL/PGSQL functions, exception blocks were not handled properly. Depending on where the exception is encountered during function execution, the improper block handling resulted in either the creation of catalog inconsistency between master and segment, or Greenplum Database issuing the following message:

The distributed transaction 'Prepare' broadcast failed to one or more segments.

22655 Locking, Signals, Processes 4.3.1 Greenplum Database hung due to incorrect lock handling that caused a race condition. The lock handling issue was caused by a compiler optimization.
20924 Dispatch 4.3.1 For some queries that contained a window function and that executed on both the master and segments, the query would hang when executed from an ODBC/JDBC client.
21899 Backup and Restore 4.3.1 When performing an incremental backup, the gpcrondump utility backed up temporary tables that existed during the time of the backup. This caused a failure when performing a restore with the gpdbrestore utility that used the incremental backup.
22293 Backup and Restore 4.3.1 Greenplum Database supports Data Domain DDOS 5.4.
22442 Loaders: gpfdist 4.3.1 The Greenplum Database Load Tools for Windows installation did not include the gssapi and auth libraries. This issue has been resolved.
19476 Client Access Methods and Tools 4.3.1 Running multiple gpload sessions simultaneously that loaded data into the same table resulted in inconsistent data in the table.
22863 DDL and Utility Statements 4.3.1 When > (greater than) was used in the CREATE OPERATOR CLASS command as an operator name, this error was returned.

operator > is not a valid ordering operator when using operator classes

22219 Query Planner 4.3.1 In certain queries that contain the median function and a GROUP BY clause, the query planner produced an incorrect plan in which some necessary columns were not projected in the operator nodes. This caused an error when trying to look up the missing columns.
22084 OS Abstraction 4.3.1 Improved handing of situations where Greenplum Database encounters segment violation errors.
17995 DDL and Utility Statements 4.3.1 In some cases, the functions pg_cancel_backend() and pg_terminate_backend() did not terminate sessions.
17773 DDL and Utility Statements 4.3.1 Greenplum Database did not properly check privileges during certain RESET ALL operations.
17481 Catalog and Metadata, DDL and Utility Statements 4.3.1 Queries on the system view pg_partitions could fail to return when DDL statements on partitioned tables were running concurrently.
15834 Loaders: Copy/ExternalTabs 4.3.1 A COPY command cancel request (Ctrl+c) followed by another COPY command and a cancel request caused the Greenplum Database session to hang. When cancel request was attempted again, a SIGSEGV error occured.
14367 DDL and Utility Statements 4.3.1 ALTER TABLE ADD COLUMN with default NULL was not supported for append-optimized tables. This syntax is now supported.
21522 Backup and Restore 4.3 The Greenplum Database utility pg_dump printed information-level messages (messages with the label [INFO]) to stderr that were not printed in previous releases. These messages were printed even when pg_dump completes without errors.

4.3 Release

Table 33. Resolved Issues in 4.3
Issue Number Category Description
21522 Backup and Restore The Greenplum Database utility pg_dump printed information-level messages (messages with the label [INFO]) to stderr that were not printed in previous releases. These messages were printed even when pg_dump completes without errors.