Greenplum 6.x version change record common manual

One . Build the relevant

Greenplum Environment construction and basic error checking

6.x edition

Greenplum Updated content of each version

Two . 6.x Version update

https://gpdb.docs.pivotal.io/6-18/admin_guide/managing/compression.html

Release 6.0.0

Release date ：2019-09-03

• Pivotal Greenplum 6.0.0 yes Greenplum Major new versions of , It includes new and changed functions .

The new function

Pivotal Greenplum 6 Including these new functions .

• PostgreSQL Core functions
• Zstandard Compression algorithm
• Loose rules for specifying table distribution columns
• Copied table data
• Resource group function
• PL/pgSQL Process language enhancements
• Greenplum 6 Concurrency improvement in
• Additional contribution modules
• PXF edition 5.8.1
• other Greenplum Database features

PostgreSQL Core functions

Pivotal Greenplum 6 Integrated from PostgreSQL edition 8.4 To version 9.4 Several new features of .

INTERVAL Data type processing

PostgreSQL 8.4 Improved text parsing INTERVAL To conform to SQL standard .INTERVAL This will change in 5.x and 6.x Output of queries using tags between versions . for example ：

$ psql
psql (8.3.23)
Type "help" for help.

gpadmin=# select INTERVAL '1' YEAR;
 interval
----------
 00:00:00
(1 row)

$ psql
psql (9.2beta2)
Type "help" for help.

gpadmin=# select INTERVAL '1' YEAR;
 interval
----------
 1 year
(1 row)

For more information , Please refer to the date / Time type .

• other PostgreSQL function
  Greenplum database 6.0 It also includes PostgreSQL These features and changes ：

• Support user-defined I/O Conversion cast .（PostgreSQL 8.4）.

• Support column level permissions (PostgreSQL 8.4).

• A table of contents pg_db_role_setting, It supports combinations for specific databases and roles (PostgreSQL 9.0) Set server configuration parameters .

• relkind In the table of contents column value pg_class Changed to match PostgreSQL 9.3 Items in .

• Support GIN Index method (PostgreSQL 8.3).

• Postgres Planner Support SP-GiST Index access method (PostgreSQL 9.2).（GPORCA Ignore SP-GiST Indexes .）

• Postgres Planner Support ordered set aggregation and mobile aggregation (PostgreSQL 9.4).

• Support jsonb data type （PostgreSQL 9.4）.

• DELETE, INSERT, and UPDATE Support WITH Clause CTE（ Common table expression ）（PostgreSQL 9.1）.

• Sorting rules support specifying sorting order and character classification behavior for column level data （PostgreSQL 9.1）.

Be careful ： Only if all columns in the query use the same collation ,GPORCA Sorting rules . If the columns in the query use different sorting rules , that Greenplum Use Postgres Planner.

Zstandard Compression algorithm

Greenplum Database 6.0 For some database operations, a pair of zstd (Zstandard) Compression support . See enable compression .

• Loose rules for specifying table distribution columns
  In previous releases , If you specify UNIQUE Constraints and DISTRIBUTED BY Clause CREATE TABLE, Then DISTRIBUTED BY Clause must be equal to column or UNIQUE The left subset of the column .Greenplum 6.x This rule has been relaxed , In order to UNIQUE Accept any subset of the column .

• This change also affects Greenplum 6.x How to choose the rule of default distribution key . If gp_create_table_random_default_distribution close （ Default ） And you do not include DISTRIBUTED BY Clause , be Greenplum Select the table distribution key according to the following command ：

• If you specify LIKEorINHERITS Clause , be Greenplum Copy the distribution key from the source table or parent table .

• If you specify a PRIMARY KEYorUNIQUE constraint , be Greenplum Select the largest subset of all key columns as the distribution key .

• If neither constraints nor constraints are specified LIKEorINHERITS Clause , be Greenplum Select the first appropriate column as the distribution key .（ Columns with geometric or user-defined data types are not suitable for Greenplum Distribute key columns .）
  Resource group function

Greenplum The database includes these new resource group features ：

MEMORY_LIMIT When you configure Greenplum Database resource group , you No need to specify a . When you specify when MEMORY_LIMIT=0,Greenplum The database will use the resource group global shared memory pool to service queries running in the Group .
When you specify when MEMORY_SPILL_RATIO=0,Greenplum The database will now use statement_mem The server configures parameter settings to identify the initial amount of query operator memory .

When used together to configure resource groups （MEMORY_LIMIT=0 and MEMORY_SPILL_RATIO=0） when , These new features provide something similar to Greenplum Memory management scheme provided by database resource queue .

And resource group MEMORY_SHARED_QUOTA、MEMORY_SPILL_RATIO and MEMORY_LIMIT The default value of property has been set to use this memory management scheme similar to resource queue , So when you first enable resource groups , Your query will run in a memory environment similar to before .admin_groupdefault_group

• Resource group Administrators group Default group

MEMORY_LIMIT	10	0
MEMORY_SHARED_QUOTA	80	80
MEMORY_SPILL_RATIO	0	0

PL/pgSQL Process language enhancements

Greenplum database 6.0 Medium PL/pgSQL Including support for the following new features ：

additional DETAIL and HINT Text to the error message thrown by the user . You can also specify SQLSTATE and SQLERRMSG The code returns when the user throws an error （PostgreSQL 8.4）.
The RETURN QUERY EXECUTE Statement specifies the query to be dynamically executed （PostgreSQL 8.4）.
Use statements for conditional execution CASE（PostgreSQL 8.4）. see also PostgreSQL Conditions in the document .
Copied table data
The CREATE TABLE Command support DISTRIBUTED REPLICATED As a distribution strategy . If this distribution policy is specified ,Greenplum The database will distribute all rows of the table to Greenplum All in the database system Segment example .

Be careful ： Hidden system columns （ctid、cmin、cmax、xmin、xmax and gp_segment_id） Cannot reference in user query of replicated table , Because they have no single 、 Explicit value .Greenplum database column does not exist An error was returned for the query .

Greenplum 6 Concurrency improvement in

Greenplum Database 6 The following concurrency improvements are included ：

Global deadlock detector ——Greenplum Earlier versions of the database were passed as UPDATE and DELETE The operation holds an exclusive table lock to prevent global deadlock . Although this strategy can prevent deadlock , But its cost is poor performance when updating concurrently .Greenplum Database 6 Including a global deadlock detector . This backend process collects and analyzes Greenplum Locks in the cluster wait for data . If the global deadlock detector determines that there is a deadlock , It will break the deadlock by canceling one or more back-end processes . By default , The global deadlock detector is disabled , And hold table level exclusive locks for table updates . When the global deadlock detector is enabled ,Greenplum The database holds row level exclusive locks and allows concurrent updates . See global deadlock detector .
BEGIN Transaction lock optimization - Greenplum database 6 Between you and business Optimize the use of transaction locks COMMIT. This facilitates highly concurrent mixed workloads .
The upstream PostgreSQL characteristic ——Greenplum 6 Including upstream PostgreSQL characteristic , Including features for fast path locks , These features reduce lock contention . This facilitates concurrent short queries and mixed workloads .
VACUUM It is easier to skip pages that cannot be locked . This lowers the vacuum and seems to be “ Get stuck ” The frequency of , This happens in VACUUM Waiting for a block to be locked for cleanup and another session to lock the block for a long time . Now? VACUUM Skip the block that it cannot lock and retry the block later .
VACUUM Recheck the visibility of the block after deleting the dead tuple . If all the remaining tuples in the block are visible to current and future transactions , Then the block is marked as all visible .
Tables that are part of the partitioned table hierarchy but do not contain data are age frozen , So they don't have to be cleaned separately , And it will not affect the calculation of the remaining transactions before the wrap occurs ID The number of . These tables include root tables and intermediate tables in the partition hierarchy , If they are additionally optimized , It also includes their associated metadata tables . This eliminates the need to clean up the root partition to reduce the age of the table , And eliminate the cleaning of all sub tables that may be unnecessary .
Additional contribution modules

Greenplum Database 6 With these additional PostgreSQL and Greenplumcontrib Modules are distributed together ：

• Automatic interpretation
• Disk quota
• Fuzzy string matching
• gp_sparse_vector
• Page check
• sslinfo
• PXF edition 5.8.1

Greenplum Database 6.0 Include PXF 5.8.1, It introduces the following new and changed features ：

PXF S3 Connectors now support the use of Amazon S3 Select Service access S3 Upper CSV and Parquet data . Please see use S3 Select stay S3 Read up CSV and Parquet data .
PXF New and upgraded libraries are bundled to provide Java 11 Support .
timestamptz Will be Parquet When data is written to an external data source ,PXF Added support for types .
PXF Now there's a reset Command to send your local PXF All in the server instance or cluster PXF The server instance is reset to the uninitialized state . See reset PXF.
PXF No longer supported DELIMITER stay CREATE EXTERNAL TABLE command LOCATIONURI It is specified in a.
other Greenplum Database features
Greenplum database 6.0 It also includes these features and 5.x Changes compared to version ：

recursive WITH Inquire about （ Common table expression ） No longer considered Beta function , It is now enabled by default . see also Pivotal Greenplum In the database document WITH Inquire about （ Common table expression ）.
VACUUM Updated to make it easier to skip pages that cannot be locked . This change should greatly reduce VACUUM While waiting for another session “ Get stuck ” The incidence of .
appendoptimizedappendonly Alias of table storage options .
new gp_resgroup_status_per_host and gp_resgroup_status_per_segment gp_toolkit View , Used on each host and / Or display resource groups on the basis of each segment CPU And memory usage .
When primary or segment mirroring is enabled , new gp_stat_replication The view contains replication statistics . The pg_stat_replication The view contains only master replication statistics .
Apply to Windows Of Greenplum In the client and loader toolkit and The program supports OpenSSL encryption gpfdists.psql
Greenplum 6 Including some with PostgreSQL 9.6 Aggregation related performance improvements .
Apply to Windows Of Greenplum Client and Loader Tools Provided in the package gpload Utilities and Greenplum Database 5 compatible .

Changed function (6.0 Start )

Greenplum database 6.0 Including the following function changes ：

Compared to previous versions ,Greenplum The performance characteristics of the database under heavy load are in version 6 Changes in . especially , You may notice that GPSS、WAL Changes related to replication and other functions have been added to the main segment I/O operation . All customers are encouraged to perform load tests with real data , To make sure that the new Greenplum 6 The cluster configuration meets their performance requirements .

• gpbackup also gprestore Stop following Greenplum Database 6 With the installation , But in Pivotal Network Provided separately on , And it can be upgraded separately from the core database installation .

• Greenplum 6 A new hop consistent hash algorithm is used to map hash data values to Greenplum paragraph . The new algorithm ensures that new segments are added to Greenplum 6 After cluster , Only those rows hashed to the new segment need to be moved .Greenplum 6 Hashes have similar characteristics to earlier Greenplum Performance characteristics of version , But it should be able to achieve faster database expansion . Please note that , The new algorithm takes more time than the previous algorithm CPU, therefore COPY stay CPU Performance may degrade on dense systems .

• Older legacy hash functions are represented as non default hash operator classes , be known as cdbhash_*_ops. From earlier than 6.0 Of Greenplum Use non default operator classes when upgrading the database . Old operator classes are compatible , But if you mix the old operator class with the new operator class , The query will need to reassign the motion .

• Server configuration parameters gp_use_legacy_hashops Controls whether to use traditional hash functions or default hash functions when creating tables defined with distributed columns .

• The system is shown in gp_distribution_policy Include relevant Greenplum More information about database tables and strategies for distributing table data across segments , Include operator classes for distributed hash functions .

• The gpcheck Utilities are no longer included in Greenplum Database 6 in .

• gpmovemirrors、gpaddmirrors And Utilities gprecoverseg Input file format for gpexpand Has been changed .

• The new file format does not use colons : As a separator , Instead, use vertical characters |. for example , In previous releases ,gpexpand A line in the input file is similar to ：

sdw5:sdw5-1:50011:/gpdata/primary/gp9:11:9:p

The updated file format is ：

sdw5|sdw5-1|50011|/gpdata/primary/gp9|11|9|p

Besides , From the line of its input file gpaddmirrors Delete prefix .mirror A line in the previous version may be similar to ：

mirror0=0:sdw1:sdw1-1:52001:53001:54001:/gpdata/mir1/gp0

The modified format is ：

0=0|sdw1|sdw1-1|52001|53001|54001|/gpdata/mir1/gp0

• Greenplum Use direct dispatch to locate usage Query for IS NULL, Similar to the query of filtering table distribution key column .

• gpinitsystem The option to specify the alternate master data directory is from change -F by -S. The -S Option no longer specifies the propagation image . Introduced a new gpinitsystem Option to specify the image configuration ：–mirror-mode={group|spread}.

• Default values for server configuration parameters log_rotation_size From 0 Change to 1GB. This will change the default log rotation behavior , In order to write more than... In the current log file 1GB Or the current log file is open 24 Open a new log file when hours .

• Default values for server configuration parameters effective_cache_size From 512MB Change to 16GB.

• The gpssh-exkeys The utility now requires that you have set no password from the primary host to all other hosts in the cluster SSH. And then run gpssh-exkeys Set no password from each host to each other host SSH.

• intelligence gpstop Shutdown behavior has changed . before , If you run gpstop -M smart（ Or just gpstop）, If there is any active client connection , The utility will exit with a message . Now? ,gpstop Wait for the current connection to complete before shutting down . If any connection is still open after the timeout period , Or if you use CTRL-C interrupt , be gpstop List the open connections and prompt to continue waiting for the connection to complete , Or perform a quick or immediate shutdown . The default timeout is 120 second , Can pass -t timeout_seconds Options to change .

• stay pg_stat_activity and pg_stat_replication In the system view , The procpid The column has been renamed to pid To match PostgreSQL 9.2 Relevant changes in .

• stay pg_proc In the system table , The proiswin The column is renamed proiswindow And reposition in the table to match pg_procPostgreSQL 8.4 System table in .

• Use SELECT DISTINCT And UNION/INTERSECT/EXCEPT Queries that no longer have to return sorted output . before , These queries always use sorting / Delete duplicate rows only . They now implement hashing to match PostgreSQL 8.4 Behavior introduced in ; This method does not produce sort output . If your application needs the sorted output of these queries , Please change the query to use explicit ORDER BY Clause . Please note that ,SELECT DISTINCT ON Never use hashes , Therefore, its behavior has not changed from the previous version .

• stay gp_toolkit Architecture , View gp_resgroup_config No more columns proposed_concurrency、proposed_memory_limit and .proposed_memory_shared_quotaproposed_memory_spill_ratio

• stay pg_resgroupcapability In the system table , The proposed Column has been deleted .

• System pg_database surface datconfig Column deleted .Greenplum The database is now in use pg_db_role_setting System table to track the configuration settings of each database and each role server （PostgreSQL 9.0）.

• Deleted pg_aggregate The system tables aggordered Column , And added several new columns to the table , To support the use of Postgres Planner (PostgreSQL 9.4) Ordered set aggregation and mobile aggregation .ALTER/CREATE/DROP AGGREGATESQL The command signature has also been updated to reflect the directory pg_aggregate change .

• System pg_authid surface rolconfig Column deleted .Greenplum The database is now in use pg_db_role_setting System table to track the configuration settings of each database and each role server （PostgreSQL 9.0）.

• When creating and changing tables with distributed columns , You can now specify hash functions for distributing data across segment instances .

• Pivotal Greenplum Database 6 from DDL (PostgreSQL 8.4) The was deleted from RECHECK Options .Greenplum It is now possible to dynamically determine whether the index operator “ lossy ”.ALTER OPERATOR FAMILYCREATE OPERATOR CLASS

• The system catalog table related to operators has been modified to support operator families 、 Compatibility and type （ Sort or search ）.

• HyperLogLog (HLL) function 、 System catalog table entries for aggregates and types are modified to have names prefixed with gp_. rename HLL Function can prevent and use HLL The outside of the Greenplum Name conflict occurred in database extension . For using built-in Greenplum database HLL Function must be updated to use the new gp_ name .

• Greenplum In previous versions “ Legacy optimizer ” It is now called Postgres Planner .

• since PostgreSQL 9.1 Introduces serializable snapshot isolation (SSI) Since mode ,Greenplum database 6.0 The transaction isolation level in has been changed to and PostgreSQL The transaction isolation level is consistent . new SSI Pattern （ Not in Greenplum Implementation in database ） Provide true serializability by monitoring concurrent transactions and rolling back transactions that may introduce serialization exceptions . Existing snapshot isolation (SI) The schema ensures that transactions are in a single... Of the database 、 Operate on consistent snapshots , But there is no guarantee of consistent results when executing a set of concurrent transactions in any given order .

• Greenplum database 6.0 It is now allowed to REPEATABLE READSQL Use keywords in statements , for example BEGIN and SET TRANSACTION.PostgreSQL 9.1 Or later SERIALIZABLE Transactions use new SSI Pattern .SERIALIZABLEGreenplum database 6.0 Transactions in fall back to REPEATABLE READ, Use SI Pattern . The following table shows Greenplum database 6.0 and PostgreSQL 9.1 For each transaction isolation level in SQL Standard compliance .

Table space

• CREATE TABLESPACE The command has changed .

This command no longer requires the gpfilespace File space created by utility .
The FILESPACE The clause has been deleted .
Added this WITH Clause to allow specifying a tablespace location for a particular segment instance .
Share the same content ID The primary mirror pair of must use the same tablespace location .

• SQL The command has new ALTER SEQUENCE Clause START [WITH] start and OWNER TO new_owner(PostgreSQL 8.4). The START Clause sets the starting value that future commands will use ALTER SEQUENCE RESTART, But the current value of the sequence will not be changed . The OWNER TO Clause to change the owner of the sequence .

• ALTER TABLESQL The command has a clause that adds system columns to the table （PostgreSQL 8.4 ）SET WITH OIDS.oid Please note that , It is strongly recommended not to put oid And Greenplum Use with database tables .

• CREATE DATABASESQL Command has new parameters ,LC_COLLATE and LC_CTYPE Specify the sort order and character classification for the new database .

• CREATE FUNCTIONSQL The order has a new keyword WINDOW, Indicates that the function is a window function rather than a normal function （PostgreSQL 8.4）.

• stay SQL Specify the index name in the command CREATE INDEX Now it's optional .Greenplum The database constructs a default index name from the table name and index column .

• In this order ,Greenplum The database parser allows between clauses and their corresponding clauses, as well as between consecutive clauses CREATE TABLE Place comma . Using this unrecorded syntax will generate a deprecation warning message .SUBPARTITION TEMPLATESUBPARTITION BYSUBPARTITION BY

• Now you need super user rights to create protocols . See create agreement .

• CREATE TYPESQL The command has a new clause , It is from an existing type （PostgreSQL 8.4）LIKE=type Copy the representation of the new type （INTERNALLENGTH、PASSEDBYVALUE、ALIGNMENT and ）.STORAGE

• GRANTSQL The command has a new syntax to grant truncation 、 External data wrappers and external data servers (PostgreSQL 8.4) Authority .

• SQL Command has LOCK One option ONLY keyword （PostgreSQL 8.4）. After designation , The table will be locked , Without locking any tables inherited from it .

• Use outside the transaction LOCK table The statement will be in the Greenplum database 6.0 An error was raised in . In previous releases , Execute statement , Although it is only useful when executed within a transaction .

• and SQL Command support SQL 2008SELECT And grammar ( PostgreSQL 8.4). These clauses provide another syntax to restrict the results returned by the query .VALUESOFFSETFETCH

• The FROM Clauses can be from SELECT Omit from the command , but Greenplum The database no longer allows queries to omit this FROM Clause and reference the database table .

• and SQL The keyword has been changed from reserved to unreserved ,ROWS also RANGE It can be used as a table or column name without reference .

• stay Greenplum 6 in , Queries on external tables with descendants will recurse into the post representation by default . This is the same as before Greenplum Changes in database versions , It never recurses into descendants . To get Greenplum 6 Previous behavior in , You must include ONLY Keyword to restrict the query to the parent table .

• Default values for server configuration parameters optimizer_force_multistage_agg From change true by false.GPORCA Now by default, a one-stage or two-stage aggregation plan will be selected according to the qualified aggregation with different cost scalars .

• SQL Command has TRUNCATE One option ONLY keyword （PostgreSQL 8.4）. When specified , The table will be truncated without truncating any tables inherited from it .

• Command line createdb Utilities have new options -l( --locale)、–lc-collate and ,–lc-ctype Used to specify the database (PostgreSQL 8.4) Locale and character classification .

• And utilities have a new option pg_dump, Instructs the utility to execute after connecting to the database and before starting a dump or restore operation (PostgreSQL 8.4).pg_dumpallpg_restore–role=rolenameSET ROLE rolename

• And orders pg_dump That's ok pg_dumpall The utility has a new option –lock-wait-timeout=timeout（PostgreSQL 8.4）. After designation , If the utility cannot acquire the shared table lock within the specified number of milliseconds , Dump will fail , Instead of waiting indefinitely .

• And command line options have been changed from -d And utilities . Corresponding long version , And still supported . Added a new option , For the in place upgrade utility .-Dpg_dumppg_dumpall–inserts–column-inserts–binary-upgrade

• ( -w)–no-password Option has been added to pg_dump、pg_dumpall and pg_restore Utility .

• This option will be removed from the utility -D Delete in .gpexpand The extended mode will be in postgres Create in the database .

• The gpstate Utility has a new -x Options , It shows the details of the ongoing system expansion .gpstate -s If no options are specified ,gpstate It also reports whether the system expansion is in progress .

• The pg_restore The utility has a new option -j( --number-of-jobs) Parameters . This option can load data by running at the same time 、 Create indexes, create constraints and other tasks to reduce the time to recover large databases .

• The vacuumdb The utility has a new -F( --freeze) Option to freeze row transaction information .

• ALTER DATABASE Include SET TABLESPACE Change the clause of the default table space .

• CREATE DATABASE Including... For setting the collation and character classification of the new database COLLATE And options .CTYPE

• stay gp_toolkit In the pattern ,gp_workfile_ because Greenplum Database 6 Enhancement of working documents , The view has changed . Information about views , See checking disk overflow space usage .gp_workfile_

• Server configuration parameters gp_workfile_compress_algorithm Has been changed to gp_workfile_compression. When working file compression is enabled ,Greenplum Database usage Zstandard Compress .

• Oracle Compatibility Functions Now as an extension in Greenplum Available in the database , be based on https://github.com/orafce/orafce Upper PostgreSQL orafce project . You don't need to perform SQL Script to install compatibility features in the database , But to perform SQL command CREATE EXTENSION orafce.Greenplum database 6.0 orafce The extension is based on orafce 3.7 edition . of Greenplum Database compatibility functions and PostgreSQL orafce Expand the information about the differences between , see also Oracle Compatibility functions .

• Greenplum Database 6 It supports specifying table columns of data types citext Is the distribution key .

• Greenplum Database 6 Provides a client and loader toolkit , You can download and install it on the client system . Former Greenplum Version provides separate client and loader packages . of Greenplum 6 More information about client packages , Please refer to the client tools in the platform requirements document .

• Greenplum database 6 Including from PostgreSQL And from Greenplum Of contrib modular . Most of these modules are now packaged as extensions , You can use CREATE EXTENSION name Command in Greenplum Register extensions in . About in Greenplum database 6 More information about registering modules in , See installing the additional modules provided .contrib

• When you enable Greenplum Database high availability , The main segment instance is based on the pre write log (WAL) The stream replication of is synchronized with the mirrored segmented instance . See the overview of segment mirroring .

• gp_stat_replication The view contains replication statistics when primary or segment mirroring is enabled .

• In previous releases , The segment image adopts the physical file replication scheme .

• stay gp_segment_configuration In the table ,replication_port deleted .datadir This column has been added to show the segment instance data directory . Column mode The value is now s（ Sync ） or n（ Out of sync ）. Use gp_stat_replication View determines the synchronization status .

• be used for Windows Of Greenplum Database 6 The client and loader toolkit does not support gpfdist Program as local Windows Service running .

Deleted features

• Pivotal Greenplum Database 6.0 Removed these features ：

• The gpseginstall Utilities are no longer included . You must be in every Segment Installed on the host Greenplum Software RPM, If installed Greenplum Database software .

• The gptransfer Utilities are no longer included ; be used for gpcopy All functions provided gptransfer.

• gp_fault_strategy No longer use system tables .Greenplum The database is now in use gp_segment_configuration System table to determine whether mirroring is enabled .

• Pivotal Greenplum Database 6 Deleted gpcrondump、gpdbrestore and gpmfr Administration utility . Use gpbackup and gprestore To backup and restore Greenplum database .

• Pivotal Greenplum Database 6 No longer supported Veritas NetBackup.

• Pivotal Greenplum Database 6 Direct... Is no longer supported I/O To bypass the memory buffer in the file system cache for backup .

• Pivotal Greenplum Database 6 No longer supported gphdfs External table protocol to access Hadoop System . In version 6 Use in Greenplum platform - Extension framework (PXF) visit Hadoop. Information about using external table protocols , see also pxf:// agreement .pxf

• Pivotal Greenplum Database 6 No longer supported SSLv3.

• Pivotal Greenplum Database 6 The following server configuration parameters have been deleted ：

gp_analyze_relative_error
gp_backup_directIO
gp_backup_directIO_read_chunk_mb
gp_connections_per_thread
gp_enable_sequential_window_plans
gp_idf_deduplicate
gp_snmp_community
gp_snmp_monitor_address
gp_snmp_use_inform_or_trap
gp_workfile_checksumming

• Greenplum database 6 Deleted in Unrecorded gp_cancel_query() Functions and configuration parameters gp_cancel_query_print_log and .gp_cancel_query_delay_time
• The string_agg(expression) The function has changed from Greenplum 6 Delete in . This function concatenates text values into a string . The string_agg(expression, delimiter) Functionality is still supported .
• Pivotal Greenplum Database 6 Configuration is no longer supported Greenplum The database system triggers when certain database events occur SNMP（ Simple network management protocol ） The ability to alert or send email notifications to system administrators . Use Pivotal Greenplum Command Center Alarm to detect and respond Greenplum Events that occur in the system .
• Pivotal Greenplum Database 6 Deleted the gpfilespace Utilities . The CREATE TABLESPACE The command no longer requires the file space created by the utility .
• Pivotal Greenplum Database 6 No longer automatically from deprecated timestamp formats YYYYMMDDHH24MISS Convert text . In the past Greenplum The format cannot be explicitly resolved in the database version .PostgreSQL 9.4 The format is not supported .

for example , This command is on the Greenplum Database 6 Error returned in . In previous releases , Return timestamp .

select to_timestamp('20190905140000');

• stay Greenplum Database 6 in , This command returns a timestamp .

select to_timestamp('20190905140000','YYYYMMDDHH24MISS');

• Pivotal Greenplum Database 6 from 、 and This option has been removed from the utility –ignore-version.pg_dumppg_dumpallpg_restore

• The gpcheck Utilities are no longer included .

Deprecated features

• Deprecated features will be in Greenplum Delete in the future major version of the database .VMware Tanzu Greenplum 6.x Abandoning ：

• gpsys1 Utilities .
  The analzyedb Options –skip_root_stats（ since 6.2 It has been abandoned ）.

• If this option is specified , A warning will be given , Indicate that this option will be ignored .

• Server configuration parameters gp_statistics_use_fkeys（ since 6.2 It has been abandoned ）.

• Server configuration parameters gp_ignore_error_table（ since 6.0 It has been abandoned ）.

• To avoid Greenplum Database syntax error , Please set the value of this parameter to true When you run the execution contains the now deleted Greenplum database 4.3.x Clause CREATE EXTERNAL TABLE The command of the application .COPYINTO ERROR TABLE

• Specify... In the order => Is the name of the operator CREATE OPERATOR（ since 6.0 It has been abandoned ）.

• Greenplum External table C API（ since 6.0 It has been abandoned ）.

• Encourage the use of this API Any developer using the new Foreign Data Wrapper API Instead of it .

• Place commas in SUBPARTITION TEMPLATE Clause corresponds to SUBPARTITION BY Between clauses , And the continuity in the command SUBPARTITION BY Between clauses CREATE TABLE（ since 6.0 It has been abandoned ）.

• Using this unrecorded syntax will generate a deprecation warning message .

• Timestamp format YYYYMMDDHH24MISS（ since 6.0 It has been abandoned ）.

• This format is in the previous Greenplum Cannot be explicitly resolved in the database version , And in PostgreSQL 9.4 Not supported in .

• createlang And Utilities （ since 6.0 Since droplang Abandoning ）.

• The system view （ since 6.0 Since pg_resqueue_status Abandoning ）.

• Please switch to gp_toolkit.gp_resqueue_status View .

• Use and Command when creating a temporary table GLOBALand Modifier （ since 6.0 It has been abandoned ）.LOCALCREATE TABLECREATE TABLE AS

• These keywords are used for SQL Standards compatibility , But in Greenplum Invalid in database .

• Use... When creating or changing tables WITH OIDS or oids=TRUE Distribute OID System column （ since 6.0 It has been abandoned ）.

• Allow super users to specify SQL_ASCII Code regardless of locale （ since 6.0 It has been abandoned ）.

• When data incompatible with locale encoding is stored in the database , This choice may cause the wrong behavior of string functions .

• @@@ Text search operator （ since 6.0 It has been abandoned ）.

• This operator is currently synonymous with the operator @@.

• The bracketed syntax of the list of options in the command VACUUM（ since 6.0 It has been abandoned ）.

• This syntax requires that the options of the commands be specified in a specific order .

• plain pgbouncer Authentication type ( auth_type= plain)（ since 4.x It has been abandoned ）.

3、 ... and . Common query

Collection ： System table query statement

Four .PostgreSQL manual

postgresql Manual address

• Memory

Parameters	describe	Use
shared_buffers (integer)	Set the amount of shared memory buffers that the database server will use	< 40%* RAM + Cache mechanism
huge_pages (enum)	Enable / Disable the use of huge memory pages try（ Default ）、on、 and off	When huge_pages Set to try when , Server will Try using giant pages , If it fails, it will switch back to the normal allocation . If set to on, Failure to use giant pages will prevent the server from starting . If set to off, You won't use giant pages .
max_prepared_transactions (integer	The setting can be in "prepared" Maximum number of transactions in state , This parameter can only be set when the server is started	If you are using preparatory Services , You will want to put max_prepared_transactions At least set to max_connections The same big as that , Therefore, each session can have a preparatory transaction to be processed .
work_mem (integer)	Specify the amount of memory used by internal sort operations and hash tables before writing to temporary disk files .	Keep this in mind when selecting this value .ORDER BY、DISTINCT Sorting operations are used for both merge and join . Hash join 、 Hash based aggregation and hash based aggregation IN Hash tables are used in subquery processing .
maintenance_work_mem (integer)	Specify in the maintenance operation （ for example VACUUM、CREATE INDEX and ALTER TABLE ADD FOREIGN KEY） Used in Maximum amount of memory	Set this value higher than work_mem Much larger is safe . Larger settings can improve the performance of cleaning up and restoring database dumps

• When running a backup server , This parameter must be at least as large as that on the primary server . otherwise , The backup server will not execute queries |

• shared_buffers : In order to distribute the processing of writing a large amount of new or changed data over a long period of time , shared_buffers Bigger Setting usually requires that max_wal_size Also increase accordingly .

• huge_pages At present , Only Linux This feature is supported on . On other systems, this parameter is set to try when , It will be ignored .

• maintenance_work_mem (integer) Note that when automatic cleaning is running , May allocate up to this memory autovacuum_max_workers times , So be careful not to set the default value too high . By setting it independently autovacuum_work_mem It may help to control this situation .