E.17. Postgres Pro Enterprise 10.1.1

Release date: 2018-02-02

E.17.1. Overview

This release is based on Postgres Pro Enterprise 9.6.6.3, and PostgreSQL 10.1. It includes all the new features introduced in PostgreSQL 10, as well as bug fixes implemented in PostgreSQL 10.1. For their detailed description, see PostgreSQL 10 Release Notes and PostgreSQL 10.1 Release Notes, respectively. As compared with Postgres Pro Enterprise 9.6.6.3, this version adds the following new features:

  • pg_shardman. This experimental extension enables sharding, aiming for scalability and fault tolerance with transaction support. It provides best results for OLTP workloads. (See pg_shardman.)

  • in-memory. This extension enables you to store data in Postgres Pro Enterprise shared memory. (See Section F.24.)

  • vops. This extension enables vertical data model for Postgres Pro Enterprise. It can speed up OLAP queries with filtering and aggregation by times. (See Section F.70.)

  • Declarative syntax for partitioning uses pg_pathman as the default partitioning engine. (See Section F.40.2.6.)

  • SCRAM-SHA-256 algorithm implementation for password encryption is replaced with the one introduced in vanilla PostgreSQL. (See Section 20.3.2.)

  • Failover algorithm for libpq connections is replaced with vanilla PostgreSQL implementation.

  • Unified the structure of binary installation packages across all Linux-based distributions. The new package structure differs from that of vanilla PostgreSQL, but offers the following benefits:

    • You do not need to worry about installation specifics on different Linux distributions when migrating between different supported Linux systems. Postgres Pro Enterprise is now installed to the /opt/pgpro/ent-10 directory, while the default database is created in the /var/lib/pgpro/ent-10/data directory.

    • You can now fully control the default database creation on all Linux distributions. If you install the postgrespro-ent-10 package, it deploys all the Postgres Pro Enterprise packages required for your system, creates the default database, and configures the server in a fully automated way. If you install individual packages, you need to configure Postgres Pro Enterprise yourself. In this case, you have to manually initialize the database cluster and start the server, as well as configure automatic server startup if required.

    • You can install Postgres Pro Enterprise 10 side by side with other PostgreSQL-based products for seamless migration or parallel work. If you are using individual packages, your current system configuration will be preserved, so you have to manually configure Postgres Pro Enterprise, as explained in Section 17.1.3. Do not install the postgrespro-ent-10 package on the same system with other PostgreSQL-based products to avoid conflicts.

    Note that postgrespro-common and postgrespro-client-common packages are no longer available on Debian-based systems. For details, see Section 17.1.

The following features have been ported from Postgres Pro Enterprise 9.6.6.3:

  • aqo extension for adaptive query optimization. (See Section F.3.)

  • multimaster and referee extensions. (See Section F.32 and Section F.56, respectively.)

  • Extension pg_hint_plan has been added (See Section F.39)

  • pgpro_scheduler (See pgpro_scheduler.)

  • pg_tsparser text search parser. (See Section F.50.)

  • pg_wait_sampling extension for sampling-based statistics of wait events. (See Section F.53.)

  • RUM index, based on GIN. (See Section F.57.)

  • Timeout for idle sessions on the server side. (See idle_session_timeout.)

  • Page-level compression (CFS). (See Chapter 32.)

  • Autonomous transactions support. (See Chapter 16.)

  • Support for relocatable tables (See pg_transfer)

  • Transaction IDs on 64-bit platforms use 64-bit data type

  • Consistent reads on standby servers. (See WAITLSN.)

  • pg_repack utility. (See pg_repack.)

  • K-nearest neighbor (k-NN) algorithm support for B-tree, GiST, and SP-GiST index types. (See Section 11.13.)

  • mamonsu monitoring service, which is implemented as a Zabbix agent. (See mamonsu.)

  • Handling transactions with serialization or deadlock errors in pgbench.

  • You can specify icu or libc as the provider for the default collation when initializing a database cluster or creating a database. By default, the icu provider is used for all locales except C and POSIX. For details, see Section 23.2.2.

  • In B-tree index, duplicate keys are now stored only once, with item pointers located in posting lists. It can significantly reduce the index size for workloads where the same key appears multiple times.

Core patches from Postgres Pro Standard 9.6 have been applied:

  • Covering indices patches (For details, see the INCLUDE clause description in CREATE INDEX.)

  • Fixes to win32 build system

  • Added pgpro_version SQL function and appropriate defines into pg_config.h

  • Integrated PTRACK patch

  • For Windows version of psql, enabled support for command-line editing using libeditline

The following modules and utilities have been ported from Postgres Pro Standard 9.6:

E.17.2. Migration to Version 10

To migrate from PostgreSQL or Postgres Pro Standard, you are recommended to perform a dump/restore using pg_dumpall. pg_upgrade utility can only be used when upgrading from Postgres Pro Enterprise. Before migration, make sure to install the latest available minor version of your product.

When migrating to Postgres Pro Enterprise 10, do not use the postgrespro-ent-10 package to avoid conflicts. Use the individual packages instead. In this case, server autostart needs to be enabled manually, if required.

Starting from Postgres Pro Enterprise 10, you can specify the provider for the default collation when initializing the database cluster or creating the database, as explained in Section 23.2.2. You must take it into account when upgrading to this release to avoid breaking indexes and constraints.

Important

For PostgreSQL 9.5 and 9.5.1, as well as Postgres Pro 9.5.0.1 and 9.5.1.2, you cannot perform an upgrade to Postgres Pro Enterprise 10 directly. If you are using one of these versions, upgrade your installation to an intermediate version first, such as Postgres Pro 9.5.2.1.

Similarly, if you are running Postgres Pro 9.6.10.x on Windows, you have to upgrade it to version 9.6.11.1 or higher.

When you are using pg_dumpall to perform the upgrade, Postgres Pro Enterprise uses the collation provider specified with the initdb command for the new cluster. In this case, indexes are rebuilt automatically. To avoid issues with collation-dependent constraints, you are recommended to use libc provider when upgrading from vanilla PostgreSQL, and omit the provider when upgrading from a previous version of Postgres Pro, unless you have databases with non-C and non-POSIX collation settings. For such databases, you should do the following:

  • If the new cluster is initialized with locales other than C or POSIX, and the database has an single-byte encoding, set LC_COLLATE to '@libc' for this database.

  • If the new cluster is initialized with C or POSIX locale settings and the database has a multi-byte encoding, set LC_COLLATE to '@icu' for this database.

Note

If the previous Postgres Pro installation contained any indexes or constraints depending on collations other than the default collation of the database, C, or POSIX in databases with multibyte encodings, such databases could contain some data that violated the specified constraints and made indexes inconsistent. On Windows, this situation can also happen if the database with a multibyte encoding contained any indexes or constraints depending on the default collation with a verbose name, such as "Russian_Russia[.encoding]" or "English_United States[.encoding]". In such cases, you can only use pg_upgrade to upgrade to Postgres Pro 10, as a dump/restore scenario may be impossible. To resolve these issues, pg_upgrade declares such indexes and constraints invalid and creates reindex_text_indexes.sql and validate_text_contraints.sql, respectively.

If you choose to run pg_upgrade when upgrading from Postgres Pro Enterprise 9.6 versions, make sure to initialize the new database cluster with the same checksum setting as the database cluster you are migrating from. To ensure that that initdb sets the correct provider of the default collation for the new database cluster, omit the provider option to select the required collation provider automatically. In this case, libc provider will be used for databases with C and POSIX locales, as well as for all databases with single-byte encodings, while icu provider will be used for all the other cases. If pg_upgrade creates any SQL files in its current directory, run these files to complete the upgrade.