E.7. Postgres Pro Enterprise 10.20.1

Release date: 2022-03-04

E.7.1. Overview

This release is based on PostgreSQL 10.20 and Postgres Pro Enterprise 10.19.1. All changes inherited from PostgreSQL 10.20 are listed in PostgreSQL 10.20 Release Notes. Other major changes and enhancements are as follows:

  • Fixed an issue with lazy conversion from 32-bit to 64-bit transaction IDs in read-only transactions or on replicas by enforcing marking pages as subject to full-page writes to WAL. This issue could cause failures to synchronize conversion results to other replicas.

  • Fixed an issue with *.cfm files that were unexpectedly left in compressed tablespaces after dropping a table.

  • Fixed the out-of-memory (OOM) killer settings in system startup files. Previously the OOM score adjustment value was set for all Postgres Pro processes rather than for postmaster only, so when Postgres Pro exhausted all RAM, the OOM killer could start killing irrelevant processes, such as sshd.

  • Ended support for outdated operating systems Debian 8 and Astra Linux Smolensk 1.5.

  • Upgraded mamonsu to version 3.4.0. Notable changes are as follows:

    • Added a new timeout parameter to the [zabbix] section of the mamonsu configuration file. It allows you to set maximum time to wait while connecting to the Zabbix server.

    • Added a new mamonsu dashboard upload template_name command, which can be used to upload a Zabbix dashboard with the mamonsu metrics to a template on the Zabbix server version 6.0 or higher.

  • Upgraded pg_probackup to version 2.5.5, which provides the following new features and bugfixes as compared to the previous included version 2.5.3:

    • Added the --checkunique option to the checkdb command to work together with --amcheck and verify unique constraints during logical verification of indexes when the amcheck extension is installed in the database and its version supports the verification of unique constraints. Check the amcheck documentation for whether this verification is supported.

    • Fixed an issue that could occur when the source database was not on the timeline 1 and the destination database did not fall behind: probackup sanity checks on catchup timeline history failed with the error Destination is not in source timeline history.

    • Fixed the behavior of catchup in DELTA and PTRACK modes. The fixed issue could occur when pg_probackup operated remotely via SSH and --destination-pgdata was the same as --source-pgdata, caused corruption of the source instance (at least the global/pg_control, global/pg_filenode.map and base/*/pg_filenode.map files got deleted) and resulted in the error Could not open file "/pgwal/test/global/pg_control" for reading: No such file or directory.

E.7.2. Migration to Version 10.20.1

If you are upgrading from a Postgres Pro Enterprise release based on the same PostgreSQL major version, it is enough to install the new version into your current installation directory.

Migration of a cluster with tables larger than 16Tb to Postgres Pro Enterprise versions 11 or 10 is not supported.

If you have previously migrated to Postgres Pro Enterprise 10.17.1 or lower, you must run the REINDEX command to rebuild GIN indexes.

While functions numeric_eq, numeric_ne, numeric_gt, numeric_ge, numeric_lt, and numeric_le are actually leakproof, they were not marked as such in Postgres Pro Enterprise 10.11.1 or lower, which could lead to incorrect query optimization. In particular, it negatively affected query execution if row-level security policy was in use. Version 10.12.1 repairs this issue for new installations by correcting the initial catalog data, but existing installations will still have incorrect markings unless you update pg_proc entries for these functions. You can run pg_upgrade to upgrade your server instance to a version containing the corrected initial data, or manually correct these entries in each database of the installation using the ALTER FUNCTION command. For example:

ALTER FUNCTION pg_catalog.numeric_eq LEAKPROOF

Starting from Postgres Pro Enterprise 10.11.1, the ICU library upgrade does not interfere with the server start. Before connecting to a database using ICU as the default collation, Postgres Pro compares this collation version to the one provided by the ICU library and displays a warning if the collation versions do not match; you may need to rebuild the objects that depend on the default collation if you think the collation change may affect the sort order of your data. To suppress these warnings, you can use the ALTER COLLATION "default" REFRESH VERSION command, as explained in ALTER COLLATION.

When upgrading from versions 10.6.2 or lower, you must run the REINDEX command to rebuild GIN, GiST, and SP-GiST indexes to fix replication issues that could be observed in these versions. You should also retake all backups for these versions if your database had such indexes.

When upgrading from versions 10.3.3 or lower, you have to rebuild GiST indexes built over columns of the intarray type, as well as indexes that use mchar or mvarchar types.

To migrate from PostgreSQL, as well as Postgres Pro Standard or Postgres Pro Enterprise based on a previous PostgreSQL major version, see the migration instructions for version 10. If you are opting for a dump/restore, make sure to use the --add-collprovider option to correctly choose the provider for the default collation of the migrated database.