For more Greenplum technology dry goods, please visitGreenplum Chinese community website

Since the official release of Greenplum 6.0, Greenplum has continued to iterate at a rapid rate of one small release per month, providing users with new features and fixes. The latest version, 6.10, was released on August 10. Here’s a look at some of the surprises that Greenplum brings in releases 6.8 through 6.10.

Summary of New features

Greenplum 6.10.1

Fix bug list

  • Interconnect agent bug fixed, enabling interconnect agent function.

  • Fixed bug WHERE DELIMITER is set to ‘OFF’ and ‘O’ is treated as DELIMITER.

Greenplum 6.10.0

New features

– New Interconnect mode: proxy mode is supported. Proxy mode consumes less connection and port resources than TCP mode, and has better performance and fault tolerance than UDPIFC mode. To use the proxy mode, set the following parameters:

  • Specify the port to use in gp_interconnect_proxy_addresses. This port must be available on maser, Standby master, primary, and mirror addresses

  • Set gp_interconnect_type to proxy

– The max_SLOt_WAL_keep_size parameter is supported. This parameter sets the maximum WAL file size for each segment. The unit is MB.

– Supports the gp_add_column_inherits_table_setting parameter, which determines whether the new column inherits the with definition of the AO table when adding columns to the AOCO table.

– New parameter gp_fts_replication_attempt_count is supported to set the maximum number of times a connection can be created at FTS.

– gpload Supports parameter FILL_MISSING_FIELDS, which allows empty fields at the end of a row to be automatically filled in during data loading.

– GPCheckCAT supports a new test mode, aoSEG_table, which can be used to check whether partition information of AOCO table is consistent with pg_attribute.

– pgbouncer is upgraded to version 1.13

– Upgrade PXF to 5.14

– GPSS is upgraded to version 1.4.1

Fix bug list

  • Fixed a bug that caused concurrent ANALYZE operations to consume too much memory

  • Fixed a bug where some queries produced large numbers of temporary files and large numbers of Hashjoins

  • Fixed a bug where gprecoverseg would hang when restoring new segment nodes in some cases

  • Fixed “unrecognized path type 106” error when some queries join partitioned tables

  • Fixed a bug where the Planner sometimes handled volatile functions incorrectly

  • Fixed a bug where views created with ‘CASE WHEN (arg1) IS NOT DISTINCT FROM (arg2)’ could NOT be returned correctly by pg_get_viewdef().

  • Fixed a bug with segment errors in some hash aggregate queries

  • GP 6.10 no longer allows you to create replicated tables while inheriting from other tables

  • Fixed “Too many distributed transactions for snapshot” error when there are multiple distributed transactions.

  • Fixed a bug where the gp_ENABLE_global_deadlock_detector parameter could not be used correctly

  • Fixed a bug where GPExpend could not properly process partitioned table children as external tables

  • Fixed gp_RESgroup_status_per_host false CPU usage

  • Fixed a bug that caused PANIC when tablespace was not checked

Greenplum 6.9.1

Fix bug list

  • The ODBC driver is upgraded to 7.16.359, which fixes SUSE compatibility problems.

  • Analyse does not work correctly when the outer table is a subtable of a partitioned table.

Greenplum 6.9.0

New features

– GPSS was upgraded to 1.4.0

– PXF upgrade to 5.13, and PXF provides a new installation tool that can be downloaded and used separately.

Fix bug list

  • Fixed bug where WAL would sometimes be abnormally removed during error recovery.

  • Fixed a bug where HEAP data would sometimes be lost when there are concurrent ALTER tables or ALTER domains

  • Alter TABLE no longer allows the distribution mode of partitioned table children to be set to replicated

  • Fixed a bug where some queries containing multistage aggregations returned incorrect results

  • Fixed cluster panic in some queries containing limit 1 or group by

  • Fixed an error message when the COALESCE data is full join

  • Fixed a bug where concurrent updates were not handled correctly when the distribution key moved to another Segment

  • Fixed a bug where resource groups could not move a query to another group with memory_limit=0 via pg_resgroup_move_query()

  • Fixed bug where restarting primary would cause some segment processes to hang

Greenplum 6.8.1

New features

– Upgrade postGIS to PostGIS-2.5.4 + Pivotal.2

Fix bug list

  • Fixed a bug where the executor would not work when ORCA generated additional pruning plans.

  • Fixed a bug where ORCA generated an incorrect query plan when it included window functions and the window functions contained subqueries.

  • Fixed performance degradation when ORCA performs equivalent operations on joins.

  • Fixed the PANIC bug caused by query containing multiple UDFs with EXECUTE ON specified.

  • Fixed a bug where PSQL \ M did not display materialized views

  • Fixed a bug where gprecoverseg would modify some segment dbid when fixing

Greenplum 6.8.0

New features

– GPSS is upgraded to version 1.3.6

– The -i parameter of gpinitSystem supports more node definition formats

– Upgrade PXF to 5.12

– PL/Container Upgraded to 2.1.2

– Support to move queries to another resource group

– Support GPCC6.2 and later data collection extensions

Fix bug list

  • ORCA no longer prints internal warnings when rolled back to Postgres’ optimizer

  • Fixed a bug that could cause lock information to be corrupted in shared memory

  • Fixed a bug where ALTER TABLE could not change the data distribution strategy of partitioned table subtables

  • Fixed a bug where ORCA did not properly use statistics for partition clipping in some cases

  • Fixed the hang bug when executing functions containing SQL statements in some cases

  • Fixed a bug where TCP interconnect was slow in some cases

  • GPORCA enhances the ability to partition clipping of text-type data

  • Fixed a bug where COPY from would crash when there was incorrect data

  • Fixed a bug where ‘code for Hash MD5 was not found’ was reported when FIPS was enabled

  • Fixed a bug where gpinitSystem incorrectly used external host names for DNS

Early review of Green Plum News

Summary of new features in Greenplum 6.1-6.7 releases

Greenplum is introduced

Greenplum is the world’s first open source, multi-cloud big data analysis platform and the only open source database in Gartner’s top ten classic and real-time data analysis products in 2019. According to the China Academy of Information and Communications Technology, 43 percent of the 14 distributed databases reviewed by the Chinese academy of Information and Communications Are based on Greenplum.

Greenplum focuses on OLAP system data engine development, and the release of version 6.0 gives Greenplum excellent support for OLTP and mixed loads. Greenplum has a world-class research and development team working at the cutting edge of high performance computing and database systems, aiming to provide high performance superdata engines for global customers and bring powerful parallel computing power to the field of large-scale data warehouse analysis.