Release date: October 14, 2026

WarehousePG Backup and Restore 1.34.1 includes bug fixes.

Note

Before you upgrade, see Upgrading for changes that might affect existing backups and scripts.

Bug fixes

• Fixed tables changed by concurrent DDL, such as ALTER TABLE ... SET WITH (REORGANIZE=true), TRUNCATE, or DROP, being backed up empty or with wrong data, or failing the backup. With --leaf-partition-data or --incremental, an append-optimized table could fail the backup with relation "pg_aoseg.pg_aoseg_<n>" does not exist. Otherwise the table was backed up empty without an error. gpbackup now takes a new snapshot and retries, up to three snapshots by default. To change the number of snapshots, see WHPGBACKUP_SNAPSHOT_ATTEMPTS. Tables that still change are left out of the data backup with a warning and listed under data not backed up: in the backup report.
• Fixed DISTRIBUTED COORDINATOR ONLY tables being backed up as 0 rows, with no error, on WarehousePG 7.4 and later. Their data is now backed up and restored through the coordinator. Custom storage plugins must now also accept backup_data and restore_data calls on the coordinator for these tables, with content -1 and the table's OID.
• Fixed an empty --plugin-config value being accepted, which made gpbackup write the backup to the segments' local disks and report success. An empty value is now an error in gpbackup, gprestore, gpbackup delete-backup, and gpbackup delete-backups-before. Restoring a plugin backup without --plugin-config now reports that the flag is required instead of a missing local directory.
• Fixed gprestore failing when the local backup directory was missing on the coordinator or a segment host while restoring through a storage plugin, for example on a rebuilt host. gprestore now creates the directories, which also fixes --resize-cluster restores from plugin storage.