Release notes for CloudNativePG 1.20

History of user-visible changes in the 1.20 minor release of CloudNativePG.

For a complete list of changes, please refer to the commits on the release branch in GitHub.

Version 1.20.2

Release date: July 27, 2023

Enhancements:

  • New logs command in the kubectl plugin, to retrieve or follow the logs of all pods in a cluster (#2375)
  • Add support for specifying priorityClassName in pods, helping Kubernetes make scheduling decisions (#2043)
  • Add a metric and status field to monitor node usage by a CloudNativePG cluster (#2257)
  • Various enhancements to the documentation:
    • Add troubleshooting instructions relating to hugepages (#1390)
    • Extend the FAQs page (#2344)

Technical enhancements:

  • Add a check at the start of the restore process to ensure it can proceed; give improved error diagnostics if it cannot (#2419)
  • Improve handling of non-expiring passwords in managed roles (#2334)

Fixes:

  • Ensure the logic of setting the recovery target matches that of Postgres (#2460)
  • Prevent taking over service accounts not owned by the cluster, by setting ownerMetadata only during service account creation (#2462)
  • Ensure correct permissions of the PGDATA directory for initdb and restore (#2384)
  • Prevent a possible crash of the instance manager during the configuration reload (#2393)
  • Prevent the LastFailedArchiveTime alert from triggering if a new backup has been successful after the failed ones (#1751)
  • Prevent services from targeting non-instance pods (#2336)

Security:

  • Updated all project dependencies to the latest versions

Version 1.20.1

Release date: June 12, 2023

Enhancements:

  • Add the snapshot command to the cnpg plugin to create a consistent cold backup of the cluster from a standby using the Kubernetes VolumeSnapshot standard resource (#1960)
  • First implementation of recovery from a set of CSI VolumeSnapshot resources via the .spec.bootstrap.recovery.volumeSnapshot stanza (#1960)
  • Add pg_failover_slots to managed extensions (#2057)
  • Improved Grafana dashboard with updated instructions in the documentation and the quickstart guide (#1916)
  • Introduce the schemaOnly option in the import stanza, to avoid exporting and importing data when you bootstrap a new Postgres Cluster from one or more existing databases (#2234)
  • Add support for TopologySpreadConstraints to manage scheduling of instance pods (#2202)
  • Add PodMonitor support to the Poolerfor PgBouncer (#2034)
  • Add option to override the default Kubernetes scheduler (#2013)
  • Allow configuration of deployment strategy of a Pooler resource (#1983)
  • Update default PostgreSQL version to 15.3 (#2022)
  • Use PgBouncer 1.19 by default (#2018)

Technical enhancements:

  • Updated k8s kind tested versions (#2054)
  • Declarative roles should ignore passwords if not set, easing management of previously existing roles (#2029)
  • Use separate transactions to reconcile role credentials. Before this patch, the operator would revert the synchronization of all roles if one failed (#2004)
  • Ensure fencing is removed during cluster restore (#1987)
  • Improve logging when deleting Pods (#2136)

Fixes:

  • Fix unbound variable with k3s engine which could prevent setup on K3’s (#2157)
  • Report the correct PG version in the metrics (#2126)
  • Use the correct walStorage key in the documentation (#2140)
  • Halt reconciliation when the operator cannot connect with the instances, and provide a clear diagnostic on such occasions. This will help clarify cases where network issues obstruct normal operation of CloudNativePG (#2145), (#2233), and (#2242)

Version 1.20.0

Release date: April 27, 2023

Important changes from previous versions

CloudNativePG 1.20 introduces some changes to the default behavior of a few features for newly created Cluster resources, compared to previous versions of the operator. The goal of these changes is to improve the resilience of a Postgres cluster out of the box through convention over configuration. For clusters with one or more replicas:

  • Backup from standby is now enabled by default, unless target is explicitly set to primary
  • Restart of the primary is now the default method to complete the unsupervised rolling update procedure (primaryUpdateMethod defaults to restart, unless explicitly set to switchover)

For further information, please refer to the "Installation and upgrades" section.

Features:

  • Declarative role management: introduce the managed.roles stanza in the Cluster spec to provide full lifecycle management of database roles, by providing an abstraction to the related DDL commands in PostgreSQL, such as CREATE ROLE and ALTER ROLE (#1524, #1793 and #1815)
  • Declarative hibernation of a PostgreSQL cluster: introduce a new annotation called cnpg.io/hibernation to declaratively hibernate a PostgreSQL cluster by deleting all pods and keeping the PVCs only; the feature also implements the inverse procedure (#1657)

Enhancements:

  • Improve the --logs option of the report command of the cnpg plugin for kubectl to also include the previous logs where available (#1811)
  • The -any service is now disabled by default (#1755)

Security:

  • Enable customization of SeccompProfile through override via a local file (#1827)

Fixes:

  • Apply the PostgreSQL configuration provided by the user during the initdb bootstrap phase, before the server is started the first time (#1858)